The Generic Mapping Tools

PIC

Version 4.5.18

Technical Reference and Cookbook

by

Pål (Paul) Wessel

School of Ocean and Earth Science and Technology

University of Hawai’i at Mānoa

and

Walter H. F. Smith

Laboratory for Satellite Altimetry

NOAA/NESDIS/STAR

July 2018

PIC

Contents

Front page
List of Tables
List of Figures
Acknowledgments
The GMT Documentation Project
A Reminder
Copyright and Caveat Emptor!
Typographic conventions
1 Preface
 1.1 What is new in GMT 4.x?
  1.1.1 Overview of GMT 4.5.18 [Jul-1, 2018]
  1.1.2 Overview of GMT 4.5.17 [Jan-1, 2018]
  1.1.3 Overview of GMT 4.5.16 [June-25, 2017]
  1.1.4 Overview of GMT 4.5.15 [October-1, 2016]
  1.1.5 Overview of GMT 4.5.14 [Nov-1, 2015]
  1.1.6 Overview of GMT 4.5.13 [Jan-1, 2015]
  1.1.7 Overview of GMT 4.5.12 [Mar-1, 2014]
  1.1.8 Overview of GMT 4.5.11 [Nov-5, 2013]
  1.1.9 Overview of GMT 4.5.9 [Jan-1, 2013]
  1.1.10 Overview of GMT 4.5.8 [Apr-1, 2012]
  1.1.11 Overview of GMT 4.5.7 [Jul-15, 2011]
  1.1.12 Overview of GMT 4.5.6 [Mar-1, 2011]
  1.1.13 Overview of GMT 4.5.5 [Nov-1, 2010]
  1.1.14 Overview of GMT 4.5.4 [Nov-1, 2010]
  1.1.15 Overview of GMT 4.5.3 [Jul-15, 2010]
  1.1.16 Overview of GMT 4.5.2 [Jan-15, 2010]
  1.1.17 Overview of GMT 4.5.1 [Sept-20, 2009]
  1.1.18 Overview of GMT 4.5.0 [July-15, 2009]
  1.1.19 Overview of GMT 4.4.0 [Feb-15, 2009]
  1.1.20 Overview of GMT 4.3.1 [May-15, 2008]
  1.1.21 Overview of GMT 4.3.0 [May-1, 2008]
  1.1.22 Overview of GMT 4.2.1 [October-10, 2007]
  1.1.23 Overview of GMT 4.2.0 [April-1, 2007]
  1.1.24 Overview of GMT 4.1.4 [Nov-1, 2006]
  1.1.25 Overview of GMT 4.1.3 [June-1, 2006]
  1.1.26 Overview of GMT 4.1.2 [May-15, 2006]
  1.1.27 Overview of GMT 4.1.1 [Mar-1, 2006]
  1.1.28 Overview of GMT 4.1 [Jan-7, 2006]
  1.1.29 Overview of GMT 4.0 [Oct-10, 2004]
2 Introduction
3 GMT overview and quick reference
 3.1 GMT summary
 3.2 GMT quick reference
4 General features
 4.1 GMT units
 4.2 GMT defaults
  4.2.1 Overview and the .gmtdefaults4 file
  4.2.2 Changing GMT defaults
 4.3 Command line arguments
 4.4 Standardized command line options
  4.4.1 Data domain or map region: The -R option
  4.4.2 Coordinate transformations and map projections: The -J option
  4.4.3 Map frame and axes annotations: The -B option
  4.4.4 Header data records: The -H option
  4.4.5 Portrait plot orientation: The -P option
  4.4.6 Plot overlays: The -K-O options
  4.4.7 Timestamps on plots: The -U option
  4.4.8 Verbose feedback: The -V option
  4.4.9 Plot positioning and layout: The -X-Y options
  4.4.10 Binary table i/o: The -b option
  4.4.11 Number of Copies: The -c option
  4.4.12 Data type selection: The -f option
  4.4.13 Data gap detection: The -g option
  4.4.14 Multiple segment data: The -m option
  4.4.15 Lat/Lon or Lon/Lat?: The -: option
 4.5 Command line history
 4.6 Usage messages, syntax- and general error messages
 4.7 Standard input or file, header records
 4.8 Verbose operation
 4.9 Program output
 4.10 Input data formats
 4.11 Output data formats
 4.12 PostScript features
 4.13 Specifying pen attributes
 4.14 Specifying area fill attributes
 4.15 Color palette tables
  4.15.1 Categorical CPT files
  4.15.2 Regular CPT files
 4.16 Character escape sequences
 4.17 Grid file format specifications
 4.18 Options for COARDS-compliant netCDF files
 4.19 The NaN data value
 4.20 GMT environment parameters
5 GMT Coordinate Transformations
 5.1 Cartesian transformations
  5.1.1 Cartesian linear transformation (-Jx-JX)
  5.1.2 Cartesian logarithmic projection
  5.1.3 Cartesian power projection
 5.2 Linear projection with polar (θ,r) coordinates (-Jp -JP)
6 GMT Map Projections
 6.1 Conic projections
  6.1.1 Albers conic equal-area projection (-Jb-JB)
  6.1.2 Equidistant conic projection (-Jd-JD)
  6.1.3 Lambert conic conformal projection (-Jl-JL)
  6.1.4 (American) polyconic projection (-Jpoly-JPoly
 6.2 Azimuthal projections
  6.2.1 Lambert Azimuthal Equal-Area (-Ja-JA)
  6.2.2 Stereographic Equal-Angle projection (-Js-JS)
  6.2.3 Perspective projection (-Jg-JG)
  6.2.4 Orthographic projection (-Jg-JG)
  6.2.5 Azimuthal Equidistant projection (-Je-JE)
  6.2.6 Gnomonic projection (-Jf-JF)
 6.3 Cylindrical projections
  6.3.1 Mercator projection (-Jm-JM)
  6.3.2 Transverse Mercator projection (-Jt-JT)
  6.3.3 Universal Transverse Mercator (UTM) projection (-Ju-JU)
  6.3.4 Oblique Mercator projection (-Jo-JO)
  6.3.5 Cassini cylindrical projection (-Jc-JC)
  6.3.6 Cylindrical equidistant projection (-Jq-JQ)
  6.3.7 Cylindrical equal-area projections (-Jy-JY)
  6.3.8 Miller Cylindrical projection (-Jj-JJ)
  6.3.9 Cylindrical stereographic projections (-Jcyl_stere-JCyl_stere)
 6.4 Miscellaneous projections
  6.4.1 Hammer projection (-Jh-JH)
  6.4.2 Mollweide projection (-Jw-JW)
  6.4.3 Winkel Tripel projection (-Jr-JR)
  6.4.4 Robinson projection (-Jn-JN)
  6.4.5 Eckert IV and VI projection (-Jk-JK)
  6.4.6 Sinusoidal projection (-Ji-JI)
  6.4.7 Van der Grinten projection (-Jv-JV)
7 Creating GMT Graphics
 7.1 The making of contour maps
 7.2 Image presentations
 7.3 Spectral estimation and xy-plots
 7.4 A 3-D perspective mesh plot
 7.5 A 3-D illuminated surface in black and white
 7.6 Plotting of histograms
 7.7 A simple location map
 7.8 A 3-D histogram
 7.9 Plotting time-series along tracks
 7.10 A geographical bar graph plot
 7.11 Making a 3-D RGB color cube
 7.12 Optimal triangulation of data
 7.13 Plotting of vector fields
 7.14 Gridding of data and trend surfaces
 7.15 Gridding, contouring, and masking of unconstrained areas
 7.16 Gridding of data, continued
 7.17 Images clipped by coastlines
 7.18 Volumes and Spatial Selections
 7.19 Color patterns on maps
 7.20 Custom plot symbols
 7.21 Time-series of RedHat stock price
 7.22 World-wide seismicity the last 7 days
 7.23 All great-circle paths lead to Rome
 7.24 Data selection based on geospatial criteria
 7.25 Global distribution of antipodes
 7.26 General vertical perspective projection
 7.27 Plotting Sandwell/Smith Mercator img grids
 7.28 Mixing UTM and geographic data sets
 7.29 Gridding spherical surface data using splines
 7.30 Trigonometric functions plotted in graph mode
8 Creating GMT Animations
 8.1 Animation of the sine function
 8.2 Examining DEMs using variable illumination
 8.3 Orbiting a static map
 8.4 Flying over topography
9 Mailing lists, updates, and bug reports
A GMT supplemental packages
 A.1 dbase: gridded data extractor
 A.2 gshhg: GSHHG data extractor
 A.3 imgsrc: gridded altimetry extractor
 A.4 meca: seismology and geodesy symbols
 A.5 mex: Matlab/Octave–GMT interface
 A.6 mgd77: MGD77 extractor and plotting tools
 A.7 mgg: GMT-MGD77 extractor and plotting tools
 A.8 misc: Miscellaneous tools
 A.9 segyprogs: plotting SEGY seismic data
 A.10 sph: spherical triangulation and gridding
 A.11 spotter: backtracking and hotspotting
 A.12 x2sys: track crossover error estimation
 A.13 x_system: track crossover error estimation
 A.14 xgrid: visual editor for grid files
B GMT file formats
 B.1 Table data
  B.1.1 ASCII tables
  B.1.2 Binary tables
  B.1.3 NetCDF tables
 B.2 Grid files
  B.2.1 NetCDF files
  B.2.2 Gridline and Pixel node registration
  B.2.3 Boundary Conditions for operations on grids
  B.2.4 Native binary grid files
 B.3 Sun raster files
C Including GMT graphics into your documents
 C.1 Making GMT Encapsulated PostScript Files
 C.2 Converting GMT PostScript to PDF or raster images
  C.2.1 When converting or viewing PostScript goes awry
  C.2.2 Using ps2raster
 C.3 Examples
  C.3.1 GMT graphics in LATEX
  C.3.2 GMT graphics in PowerPoint
 C.4 Concluding remarks
D Availability of GMT and related code
 D.1 Source distribution
 D.2 Pre-compiled Executables
E Predefined bit and hachure patterns in GMT
F Chart of octal codes for characters
G PostScript fonts used by GMT
H Problems with display of GMT PostScript
 H.1 PostScript driver bugs
 H.2 Resolution and dots per inch
 H.3 European characters
 H.4 Hints
I Color Space: The final frontier
 I.1 RGB color system
 I.2 HSV color system
 I.3 The color cube
 I.4 Color interpolation
 I.5 Artificial illumination
 I.6 Thinking in RGB or HSV
 I.7 CMYK color system
J Filtering of data in GMT
K The GMT High-Resolution Coastline Data
 K.1 Selecting the right data
 K.2 Format required by GMT
 K.3 The long and winding road
 K.4 The Five Resolutions
  K.4.1 The crude resolution (-Dc)
  K.4.2 The low resolution (-Dl)
  K.4.3 The intermediate resolution (-Di)
  K.4.4 The high resolution (-Dh)
  K.4.5 The full resolution (-Df)
L GMT on non-UNIX platforms
 L.1 Introduction
 L.2 Cygwin and GMT
 L.3 SFU and GMT
 L.4 DJGPP and GMT
 L.5 WIN32 and GMT
 L.6 OS/2 and GMT
 L.7 Mac OS and GMT
M Of colors and color legends
 M.1 Built-in color palette tables
 M.2 Labeled and non-equidistant color legends
N Custom Plot Symbols
O Annotation of Contours and “Quoted Lines”
 O.1 Label Placement
 O.2 Label Attributes
 O.3 Examples of Contour Label Placement
  O.3.1 Equidistant labels
  O.3.2 Fixed number of labels
  O.3.3 Prescribed label placements
  O.3.4 Label placement at simple line intersections
  O.3.5 Label placement at general line intersections
 O.4 Examples of Label Attributes
  O.4.1 Label placement by along-track distances, 1
  O.4.2 Label placement by along-track distances, 2
  O.4.3 Using a different data set for labels
 O.5 Putting it all together
P Special Operations
 P.1 Running GMT in isolation mode
 P.2 Using both GMT 3 and 4

List of Tables

4.1 The 17 standardized GMT command line switches.
4.2 Interval type codes.
4.3 Interval unit codes.
4.4 GMT predefined pen widths.
4.5 A few examples of pen specifications.
4.6 A few examples of fill specifications.
4.7 GMT text escape sequences.
4.8 Shortcuts for some European characters.
4.9 GMT grid file formats.
6.1 Standard parallels for some cylindrical equidistant projections.
6.2 Standard parallels for some cylindrical equal-area projections.
6.3 Standard parallels for some cylindrical equal-area projections.
B.1 Attributes of default GMT grid file in COARDS-compliant netCDF format.
B.2 GMT grid file header record. TYPE can be char, short, int, float, or double.
B.3 Structure of a Sun raster file.
B.4 Sun macro definitions relevant to raster files.

List of Figures

The four horsemen of the GMT apocalypse: Remko Scharroo, Paul Wessel, Walter H.F. Smith, and Joaquim Luis at the GMT Developer Summit in Honolulu, Hawaii during February 14–18, 2011.
4.1 Some GMT parameters that affect plot appearance.
4.2 More GMT parameters that affect plot appearance.
4.3 Even more GMT parameters that affect plot appearance.
4.4 The plot region can be specified in two different ways. (a) Extreme values for each dimension, or (b) coordinates of lower left and upper right corners.
4.5 The 30+ map projections and coordinate transformations available in GMT.
4.6 Geographic map border using separate selections for annotation, frame, and grid intervals. Formatting of the annotation is controlled by the parameter PLOT_DEGREE_FORMAT in your .gmtdefaults4 file.
4.7 Geographic map border with both primary (P) and secondary (S) components.
4.8 Linear Cartesian projection axis. Long tickmarks accompany annotations, shorter ticks indicate frame interval. The axis label is optional. We used -R0/12/0/1 -JX3/0.4 -Ba4f2g1:Frequency::,%:.
4.9 Logarithmic projection axis using separate values for annotation, frame, and grid intervals. (top) Here, we have chosen to annotate the actual values. Interval = 1 means every whole power of 10, 2 means 1, 2, 5 times powers of 10, and 3 means every 0.1 times powers of 10. We used -R1/1000/0/1 -JX3l/0.4 -Ba1f2g3. (middle) Here, we have chosen to annotate log10 of the actual values, with -Ba1f2g3l. (bottom) We annotate every power of 10 using log10 of the actual values as exponents, with -Ba1f2g3p.
4.10 Exponential or power projection axis. (top) Using an exponent of 0.5 yields a x axis. Here, intervals refer to actual data values, in -R0/100/0/1 -JX3p0.5/0.4 -Ba20f10g5. (bottom) Here, intervals refer to projected values, although the annotation uses the corresponding unprojected values, as in -Ba3f2g1p.
4.11 Cartesian time axis, example 1.
4.12 Cartesian time axis, example 2.
4.13 Cartesian time axis, example 3.
4.14 Cartesian time axis, example 4.
4.15 Cartesian time axis, example 5.
4.16 Cartesian time axis, example 6.
4.17 Cartesian time axis, example 7.
4.18 Users can specify Landscape [Default] or Portrait (-P) orientation.
4.19 A final PostScript file consists of any number of individual pieces.
4.20 The -U option makes it easy to “date” a plot.
4.21 Plot origin can be translated freely with -X -Y.
5.1 Linear transformation of Cartesian coordinates.
5.2 Linear transformation of map coordinates.
5.3 Linear transformation of calendar coordinates.
5.4 Logarithmic transformation of x-coordinates.
5.5 Exponential or power transformation of x-coordinates.
5.6 Polar (Cylindrical) transformation of (θ,r) coordinates.
6.1 Albers equal-area conic map projection
6.2 Equidistant conic map projection
6.3 Lambert conformal conic map projection
6.4 (American) polyconic projection
6.5 Rectangular map using the Lambert azimuthal equal-area projection.
6.6 Hemisphere map using the Lambert azimuthal equal-area projection.
6.7 Equal-Area (Schmidt) and Equal-Angle (Wulff) stereo nets.
6.8 Polar stereographic conformal projection.
6.9 Polar stereographic conformal projection with rectangular borders.
6.10 General stereographic conformal projection with rectangular borders.
6.11 View from the Space Shuttle in Perspective projection.
6.12 Hemisphere map using the Orthographic projection.
6.13 World map using the equidistant azimuthal projection.
6.14 Gnomonic azimuthal projection.
6.15 Simple Mercator map.
6.16 Rectangular Transverse Mercator map.
6.17 A global transverse Mercator map.
6.18 Universal Transverse Mercator zone layout.
6.19 Oblique Mercator map using -Joc. We make it clear which direction is North by adding a star rose with the -T option.
6.20 Cassini map over Sardinia.
6.21 World map using the Plate Carrée projection.
6.22 World map using the Behrman cylindrical equal-area projection.
6.23 World map using the Miller cylindrical projection.
6.24 World map using Gall’s stereographic projection.
6.25 World map using the Hammer projection.
6.26 World map using the Mollweide projection.
6.27 World map using the Winkel Tripel projection.
6.28 World map using the Robinson projection.
6.29 World map using the Eckert IV projection.
6.30 World map using the Eckert VI projection.
6.31 World map using the Sinusoidal projection.
6.32 World map using the Interrupted Sinusoidal projection.
6.33 World map using the Van der Grinten projection.
7.1 Contour maps of gridded data.
7.2 Color images from gridded data.
7.3 Spectral estimation and xy-plots.
7.4 3-D perspective mesh plot (left) and colored version (right).
7.5 3-D illuminated surface.
7.6 Two kinds of histograms.
7.7 A typical location map.
7.8 A 3-D histogram.
7.9 Time-series as “wiggles” along a track.
7.10 Geographical bar graph.
7.11 The RGB color cube.
7.12 Optimal triangulation of data.
7.13 Display of vector fields in GMT.
7.14 Gridding of data and trend surfaces.
7.15 Gridding, contouring, and masking of data.
7.16 More ways to grid data.
7.17 Clipping of images using coastlines.
7.18 Volumes and geo-spatial selections.
7.19 Using color patterns and additional PostScript material in illustrations.
7.20 Using custom symbols in GMT.
7.21 Time-series of RedHat stock price since IPO.
7.22 World-wide seismicity the last 7 days.
7.23 All great-circle paths lead to Rome.
7.24 Data selection based on geospatial criteria.
7.25 Global distribution of antipodes.
7.26 General vertical perspective projection.
7.27 Plotting Sandwell/Smith Mercator img grids.
7.28 Mixing UTM and geographic data sets requires knowledge of the map region domain in both UTM and lon/lat coordinates and consistent use of the same map scale.
7.29 Gridding of spherical surface data using Green’s function splines.
7.30 Trigonometric functions plotted in graph mode.
8.1 Animation of a simple sine function.
8.2 Animation of a DEM using variable illumination.
8.3 Orbiting a static map.
8.4 Flying over topography.
B.1 Gridline registration of data nodes.
B.2 Pixel registration of data nodes.
C.1 Examples of rendered images in a PowerPoint presentation.
C.2 PowerPoint’s “Format Picture” dialogue to set scale and rotation.
F.1 Octal codes and corresponding symbols for StandardEncoding (left) and ISOLatin1Encoding (right) fonts.
F.2 Octal codes and corresponding symbols for Symbol (left) and ZapfDingbats (right) fonts.
G.1 The standard 35 PostScript fonts recognized by GMT.
I.1 Chartreuse in GIMP. (a) Sliders indicate how the color is altered when changing the H, S, V, R, G, or B levels. (b) For a constant hue (here 90°) value increases to the right and saturation increases up, so the “pure” color is on the top right.
I.2 The 555 unique color names that can be used in GMT. Lower, upper, or mixed case, as well as the british spelling of “grey” are allowed. A4, Letter, and Tabloid sized versions of this RGB chart can be found in the GMT documentation directory.
I.3 The HSV color space.
I.4 When interpolating colors, the color system matters. The polar palette on the left needs to be interpolated in RGB, otherwise hue will change between blue (240°) and white (0°). The rainbow palette should be interpolated in HSV, since only hue should change between magenta (300°) and red (0°). Diamonds indicate which colors are defined in the palettes; they are fixed, the rest is interpolated.
J.1 Impulse responses for GMT filters.
J.2 Transfer functions for 1-D GMT filters.
J.3 Transfer functions for 2-D (radial) GMT filters.
K.1 Map using the crude resolution coastline data.
K.2 Map using the low resolution coastline data.
K.3 Map using the intermediate resolution coastline data. We have added a compass rose just because we have the power to do so.
K.4 Map using the high resolution coastline data.
K.5 Map using the full resolution coastline data.
M.1 The standard 22 cpt files supported by GMT.
M.2 The many forms of color legends created by psscale.
N.1 Custom plot symbols supported by GMT.
O.1 Equidistant contour label placement with -Gd, the only algorithm available in previous GMT versions.
O.2 Placing one label per contour that exceed 1 inch in length, centered on the segment with -Gn.
O.3 Four labels are positioned on the points along the contours that are closest to the locations given in the file fix.d in the -Gf option.
O.4 Labels are placed at the intersections between contours and the great circle specified in the -GL option.
O.5 Labels are placed at the intersections between contours and the multi-segment lines specified in the -GX option.
O.6 Labels attributes are controlled with the arguments to the -Sq option.
O.7 Another label attribute example.
O.8 Labels based on another data set (here bathymetry) while the placement is based on distances.
O.9 Tsunami travel times from the Canary Islands to places in the Atlantic, in particular New York. Should a catastrophic landslide occur it is possible that New York will experience a large tsunami about 8 hours after the event.
P.1 Example created in isolation mode

Acknowledgments

The Generic Mapping Tools (GMT) could not have been designed without the generous support of several people. We gratefully acknowledge A. B. Watts and the late W. F. Haxby for supporting our efforts on the original version 1.0 while we were their graduate students at Lamont-Doherty Earth Observatory. Doug Shearer and Roger Davis patiently answered many questions over e-mail. The subroutine gauss was written and supplied by Bill Menke. Further development of versions 2.0–2.1 at SOEST would not have been possible without the support from the HIGP/SOEST Post-Doctoral Fellowship program to Paul Wessel. Walter H. F. Smith gratefully acknowledges the generous support of the C. H. and I. M. Green Foundation for Earth Sciences at the Institute of Geophysics and Planetary Physics, Scripps Institution of Oceanography, University of California at San Diego. GMT series 3.x, 4.x, and 5.x owe their existence to grants EAR-93-02272, OCE-95-29431, OCE-00-82552, OCE-04-52126, and OCE-1029874 from the National Science Foundation, which we gratefully acknowledge.

We would also like to acknowledge the feedback we have received from many of the users of earlier versions. Many of these suggestions have been implemented, and the bug reports have been useful in providing more robust programs. Specifically, we would like to thank Michael Barck, Manfred Brands, Stephan Eickschen, Ben Horner-Johnson, John Kuhn, Angel Li, John Lillibridge, Andrew Macrae, Alex Madon, Greg Neumann, Lloyd Parkes, Ameet Raval, Georg Schwarz, Richard Signell, Peter Schimidt, Dirk Stoecker, Eduardo Suárez, Mikhail Tchernychev, Malte Thoma, David Townsend, Garry Vaughan, William Weibel, Florian Wobbe, and many others, including their advice on how to make GMT portable to a wide range of platforms. John Lillibridge provided the original example 11; Hanno von Lom helped resolve early problems with DLL libraries for Win32; Lloyd Parkes enabled indexed color images in PostScript; Kurt Schwehr maintains the Fink packages; Wayne Wilson implemented the full general perspective projection; and William Yip helped translate GMT to POSIX ANSI C and incorporate netCDF 3. The SOEST RCF staff (Ross Ishida, Pat Townsend, and Sharon Stahl) provided valuable help on Linux, web, and CGI script issues.

Honolulu, HI, College Park, MD, Cornish, NH, and Faro, Portugal, July 2018


PIC

Figure 1: The four horsemen of the GMT apocalypse: Remko Scharroo, Paul Wessel, Walter H.F. Smith, and Joaquim Luis at the GMT Developer Summit in Honolulu, Hawaii during February 14–18, 2011.

The GMT Documentation Project

Starting with GMT version 3.2, all GMT documentation was converted from Microsoft Word to LATEX files. This step was taken for a number of reasons:

  1. Having all the documentation source available in ASCII format makes it easier to access by several GMT developers working on different platforms in different countries.
  2. GMT scripts can now be included directly into the text so that the documentation is automatically up-to-date when scripts are modified.
  3. All figures are generated on the fly and included as GMT EPS files which thus are always up-to-date.
  4. It is easy to convert the LATEX files to other formats, such as HTML, SGML, PostScript, and PDF.
  5. The whole task of assembling the pieces, be it generating figures or extracting text portions from the master archive under subversion control, is automated by a makefile.
  6. Only free software are used to maintain the GMT Documentation.

Please post a New Issue from GMT home page if you find errors or inconsistencies in the documentation.

A Reminder

If you feel it is appropriate, you may consider paying us back by citing our EOS articles on GMT (and perhaps also our Geophysics article on the GMT program surface) when you publish papers containing results or illustrations obtained using GMT. The EOS articles on GMT are

The article in Geophysics on surface is

GMT includes some code supplied by others, in particular the Triangle code used for Delaunay triangulation. Its author, Jonathan Shewchuk, says

“If you use Triangle, and especially if you use it to accomplish real work, I would like very much to hear from you. A short letter or email (to jrs@cs.cmu.edu) describing how you use Triangle will mean a lot to me. The more people I know are using this program, the more easily I can justify spending time on improvements and on the three-dimensional successor to Triangle, which in turn will benefit you.”

A few GMT users take the time to write us letters, telling us of the difference GMT is making in their work. We appreciate receiving these letters. On days when we wonder why we ever released GMT we pull these letters out and read them. Seriously, as financial support for GMT depends on how well we can “sell” the idea to funding agencies and our superiors, letter-writing is one area where GMT users can affect such decisions by supporting the GMT project.

Copyright and Caveat Emptor!

Copyright ©1991 – 2018 by Paul Wessel and Walter H. F. Smith

The Generic Mapping Tools (GMT) is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation.

The GMT package is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the file LICENSE.TXT in the GMT directory or the GNU General Public License for more details.

Permission is granted to make and distribute verbatim copies of this manual provided that the copyright notice and these paragraphs are preserved on all copies. The GMT package may be included in a bundled distribution of software for which a reasonable fee may be charged.

The Generic Mapping Tools (GMT) does not come with any warranties, nor is it guaranteed to work on your computer. The user assumes full responsibility for the use of this system. In particular, the School of Ocean and Earth Science and Technology, the National Oceanic and Atmospheric Administration, the National Science Foundation, Paul Wessel, Walter H. F. Smith, or any other individuals involved in the design and maintenance of GMT are NOT responsible for any damage that may follow from correct or incorrect use of these programs.

Typographic conventions

In reading this documentation, the following provides a summary of the typographic conventions used in this document.

  1. User input and GMT or UNIX commands are indicated by using the typewriter type style, e.g., chmod +x job03.sh.
  2. The names of GMT programs are indicated by the bold, sans serif type style, e.g., we plot text with pstext.
  3. The names of other programs are indicated by the bold, slanted type style, e.g., grep.
  4. File names are indicated by the underline type style, e.g., gmt.h.

Chapter 1
Preface

While GMT has served the map-making and data processing needs of scientists since 19881, the current global use was heralded by the first official release in EOS Trans. AGU in the fall of 1991. Since then, GMT has grown to become a standard tool for many users, particularly in the Earth and Ocean Sciences. Development has at times been rapid, and numerous releases have seen the light of day since the early versions. For a detailed history of the changes from release to release, see file ChangeLog in the main GMT directory. For a nightly snapshot of ongoing activity, see the online ChangeLog page.

The success of GMT is to a large degree due to the input of the user community. In fact, most of the capabilities and options in GMT programs originated as user requests. We would like to hear from you should you have any suggestions for future enhancements and modification. Please send your comments to the user forum on the GMT home page.

1.1 What is new in GMT 4.x?

Since the release of GMT 5.x, GMT 4.x has continued to see corrections of legacy bugs and problems, but no new features will be added. Therefore, the GMT 4.x releases were mostly bug-fixes as all development is now focussed on GMT 5; this new series, released Nov-5, 2013, is distinguished by being completely restructured to allow developers access to high-level GMT modules from a variety of programming environments. Below is a brief history of the development milestones in the 4.x series. All work, including bug fixes, has now ceased for the GMT 4 series, making 4.5.18 the final release.

1.1.1 Overview of GMT 4.5.18 [Jul-1, 2018]

We have corrected one minor bug in this final release of the GMT4 series:

gmt_stat.c
: Fix issue # 1185 in t-critical calculation.

1.1.2 Overview of GMT 4.5.17 [Jan-1, 2018]

We have corrected a few minor bugs and other issues in two of the supplements:

pscontour.c
: Fix issue #1173.
meca/psmeca.c
: Fix issue #1171.
mgd77/mgd77.c
: Fix writing issue for mgd77 header.
gmt_support.c
: PostScriptwrong equations for BC d2z/dxdy for grids. Made tiny changes to ex14 PostScript and grdimage.ps test.

1.1.3 Overview of GMT 4.5.16 [June-25, 2017]

We have corrected a few bugs and other issues for individual programs in the mgd77 supplement:

mgd77/mgd77.c
: Fix format order search. Issue #1039. Set the priority of formats to match GMT5: MGD77+ first, then MGD77T, MGD77, and plain text table.
mgd77/mgd77convert.c
: Specify new search order if extension not given.
mgd77/mgd77header.c
: Added -Mt and removed unused code. Removed unrelated options from careless copy/paste via mgd77/mgd77list.c.
mgd77/mgd77info.c
: Specify new search order if extension not given.
mgd77/mgd77list.c
: Clarified usage of E77 bits in source and docs Did not use bitflags set with -Af correctly.
mgd77/mgd77path.c
: Specify new search order if extension not given.
mgd77/mgd77sniffer.c
: Replace wrong use of common options -B, -K, -P with options -M, -E, and -Z, respectively.
mgd77/mgd77track.c
: Failed to reduce time by time_inc before plotting marker. Specify new search order if extension not given.

1.1.4 Overview of GMT 4.5.15 [October-1, 2016]

We fixed some old compiler-flags that prevented compilation under recent version of Solaris and removed some copyrighted Numerical Recipes code in the meca supplement that were not even used. In addition, we made a few bug corrections for individual programs:

gmt_map.c
: Did not compute correct length of clip path for azimuth-elevation projection.
pscontour.c
: Would crash if no contours were requested.
grdtrack.c
: The -Z option did not deactivate -: for output.
surface.c
: Critical bug in which the test to determine if a Cartesian data point is “close” to a node would find most points to be close and bypass the Briggs scheme.
mgd77/mgd77list.c
: Can only use logical field tests on observed data columns.
sph/sphinterpolate.c
: Fix the vertical flipping under -Q0 mode.

1.1.5 Overview of GMT 4.5.14 [Nov-1, 2015]

Another bug-fix release with a few bug corrections for individual programs:

gmt_plot.c
: Fix issue # 662 with graph arrows for negative scales. Flipping the n/s sides array when lon and -y axes fixes issue #520.
gmt_support.c
: Fix bug in jump near Dateline in gmt_inonout_sphpol_count, affecting inside/outside spherical cases, reported in message-2219.
meca/psmeca.c
: Was unable to read any value x and y coordinates.
share/cpt/GMT_wysiwyg.cpt
: Fix typo in blue for high end [Issue #689].

1.1.6 Overview of GMT 4.5.13 [Jan-1, 2015]

Another bug-fix release. The default coastline version is now at 2.3.4. Below is the list of bug corrections for individual programs:

gmt_bcr.c
: Fixed problem with points slightly outside region causing an access violation; see issue #620.
gmt_map.c
: The map region clip path for -JE/-Je was hardwired to a full circle and did not care about -R setting. When -R...r and -JE/-Je was given the search for enclosing geographic boundaries failed when either the S or N pole was the projection’s antipole. Rect clipping with -Rw/s/e/nr could introduce stray lines when dataset has features about 180 degrees away from the user’s area. Could run into trouble clipping polygons for non-periodic maps. Added more checks in GMT_wesn_clip.
gmt_plot.c
: Left a PSL variable undefined when grdimage plotted a grid with x = longitude and y = Cartesian.
gmtmath.c
: Fix the order of when input files and -C interact.
gmtselect.c
: The -Fpolygon option might miss points because the io-machinery would sometimes set the min/max longitudes found incorrectly. The -Z option needed to pass records with NaNs.
grdfilter.c
: Fix y-shifts when -I sets smaller output spacing (issue #616).
grdmath.c
: Fix bug in grdmath’s KM2DEG operator.
psxy.c
: Fixed wrap-around issue in fault lines.
psxyz.c
: Fixed wrap-around issue in fault lines.
sample1d.c
: Fix resampling for decreasing t values.
mgd77/mgd77sniffer.c
: Needed to check that grid values used in decimate function also were within reasonable range.
x2sys/x2sys_solve.c
: Fix problems that arise when a track has less crossovers than model parameters.

1.1.7 Overview of GMT 4.5.12 [Mar-1, 2014]

Unbeknownst to many software developers, Apple changed the behavior of strcpy in Mavericks in an effort to strengthen the security of user codes. This caused ps2raster.c to crash for OS X users when it in the past had worked fine. The problem was easily corrected but meant that all those who had just upgraded to Mavericks were in a bind. This update corrects this problem, as well as several bugs. We also had to revert the behavior of the grdmath operator SDIST to again return km instead of degrees since it cannot do degrees for geodesics. By doing km it is now aligned with what PDIST and LDIST return for geographic data, and matches the behavior in GMT5. Some scripts had to be modified to handle the new behavior. If you use SDIST you may wish to check your usage. To assist users wanting spherical degrees we have added the two conversion operators KM2DEG and DEG2KM that can be used if the ELLIPSOID setting is Sphere. Finally, since GSHHG 2.3.0 has now been released we set that as the default coastline version. Below is the list of bug corrections for individual library files or programs:

gmt_grdio.c
: When building the grid command string it could exceed its maximum length if the argument became exactly the max length of the array; then the final extra zero character would be added and exceed array length. This was a rare occurrence. Thanks to Joachim Saul for pointing it out.
gmt_map.c
: Protect against longitude wrap in the Haversine and Rudoe formulae for distances. Prevent integer overflow when calculating image array size.
gmt_plot.c
: Grid crosses clipped by map boundary were badly reoriented.
gmt_proj.c
: GMT_eckert4 and GMT_eckert6 did not increment iteration counter and could in rare situations get stuck in an infinite loop.
gmt_support.c
: Allow GMT_crossover to deal with global wrapped data.
gmt_vector.c
: Add protection against introduced longitude-jumps produced by resampling in GMT_fix_up_path.
grdlandmask.c
: Failed to fill in some partial bins with no actual coastlines going through, in particular when crossing Greenwich.
grdmath.c
: Operator SDIST used spherical calculations for geodesics and ellipsoidal calculations great circles (we want the opposite). Now returning distances in km.
grdvector.c
: The new testing of -I arguments failed for some grids. Vector direction did not adjust when negative scales were used in -JX or -Jx.
ps2raster.c
: Did not check if source and destination given to strcpy were the same; this was always undefined behavior that now triggers an exit. Also added better check on the return code from the system calls in case gs or gdal_translate return with an error. Finally, make sure that image references in KML files do not carry a directory name.
pscoast.c
: The processing of -I and -N levels and pens could get out of sync so that the wrong pen was used for the specified feature.
pslib.c
: Take superscripts and subscripts into account when determining dimensions of text box. Also auto-widen paragraph width if the widest word actually is wider that chosen paragraph width.
psxy.c
: Vector direction did not adjust when negative scales were used in -JX or -Jx.
psxyz.c
: Vector direction did not adjust when negative scales were used in -JX or -Jx.
mgd77/mgd77header.c
: Fix selection of 10x10 degree identifiers.
misc/gmt2kml.c
: Make sure the default style IDs are unique for each process.
x2sys/x2sys.c
: Mishandled the assignment of segment number for each record.

1.1.8 Overview of GMT 4.5.11 [Nov-5, 2013]

Note: Due to a few technical issues we had an aborted update to 4.5.10 that was briefly released, then retracted; hence the 4.5.11-numbered version. GMT 4.5.11 is another service release with bug-fixes only. The only non-bug change was adding the latest dimensions for recent Sandwell/Smith img files that go up to 85°, and adding definition file dat.def for mgd77 ASCII DAT format to the x2sys supplement. We also had to modify the –S option in pscontour.c to address a bug. This GMT release also coincides with the latest GSHHG release version 2.2.4 which adds a few missing lakes to California and fixes an error in the Baffin Island coastline and removes skinny spikes from numerous features. Below is the list of bug corrections for individual library files or programs:

gmt_customio.c
: The magic recognition of native bit grids failed due to bad math. Wrote wrong number of bytes per record for odd-width Sun rasterfiles.
gmt_grdio.c
: Would restrict grid region in grdimage.c despite doing a global map with azimuthal projections.
gmt_io.c
: Formats for degree annotations using colons should never end in a trailing colon. Could not properly decode yyodd (no delimiter) time coordinates like 12Oct24. The GMT_import_table function checked for greenwich before assigning the input data.
gmt_init.c
: Shifted JD origin by one day (24 Nov, instead of 25 Nov).
gmt_map.c
: The oblique Mercator would get the pole on the wrong hemisphere. When -Jx is used with longitudes we must use the wesn clipping and outside functions, not the Cartesian ones. Fixed clipping problem in GMT_wesn_clip for regions larger than 180 but less than 360. GMT_grdproject_init did not handle increments that had been specified as units, e.g., -D30e.
gmt_plot.c
: Did not check for map-jumping in GMT_plot_rectangle (psxy -SJ).
gmt_proj.c
: Inverse -JR blew up at origin; now added a check. Needed to allow for minor round-off when determining if a point is beyond the horizon for -JG general perspective projection.
blockmean.c
: Did not use data near west column nodes that were off by 360 for gridline registered grids.
blockmedian.c
: Did not use data near west column nodes that were off by 360 for gridline registered grids.
blockmode.c
: Did not use data near west column nodes that were off by 360 for gridline registered grids.
filter1d.c
: Susceptible to round-off when determining t of first and last output point when -T was not given.
gmtmath.c
: The MIN and MAX operators ignored NaNs, but result should be NaN if one of the operands equal NaN. Wrong index order in rarely used SVD part of LSQFIT.
gmtset.c
: Did not write values to .gmtdefaults4 if BASEMAP_TYPE was graph or inside.
grdfft.c
: Fix normalization for std.dev of power estimate in -E.
grdimage.c
: Fix bug represented by the globalgrid.sh test script for mix of -R selections and pixel/gridline choices.
grdblend.c
: Despite geographic grids there were no check to shift a grid region by ±360 to match specified output region.
grdlandmask.c
: Did not set output as geographic after using -Jx1d.
grdmath.c
: The MIN and MAX operators ignored NaNs, but result should be NaN if one of the operands equal NaN. The XOR operator was incorrect, it is now clarified to be 0 if A == NaN and B == NaN, NaN if B == NaN, else A. Fix bug in CURV operator.
grdsample.c
: When given a -Rg grid and giving -Rg on command line, the output region became -360/0 instead of the expected 0/360.
grdtrend.c
: We messed up an interior parameter array in the 2009-10-14 fix in 4.5.2. This affected robust fits and grids with NaNs.
grdvector.c
: Did not reject vectors on far side of orthographic maps. Enforce that -Idx/dy must be multiples of grid dx/dy and abort if they are not. Before we would crash, hang, etc.
greenspline.c
: The normalization for 2-D with geographic data suffered from not checking that longitudes may be off by ±360. Needed -f in order to select -f0T input, plus it made assumptions about getting lon,lat despite not being selected. When -T was used the number of z-layers (1) was not initialized.
nearneighbor.c
: Clarify how -N works, what the defaults are, and let the minimum number of sectors default to 50% of sectors instead of a hard-wired 2.
pscontour.c
: Added -St to skip triangles whose 3 vertices are outside domain; in contrast, -S or -Sp skips all points outside domain before triangularization.
psmask.c
: Multiple, ancient bugs fixed: properly mark used edges, fix memory allocations, not report clipping if -D is used, starting point for a contour was not offset by 1/2 pixel. Was off by one in the grid index calculation.
pstext.c
: The line in -D...v was plotted on top rather than beneath box.
xyz2grd.c
: For -E, must read data as double so can properly compare with the nodata_value read as double.
meca/psvelo.cc
: Called get_trans at north pole and tried to find a point further north. Did not honor the -N setting.
mgd77/mgd77list.cc
: The azimuth written was back-azimuth, not forward. Picked id = time_column when set was 1 (custom), causing the first custom data column to be formatted as time (this is for the netCDF format files).
sph/sphdistance.c
: Make sure we visit replicated columns for gridline registered grids.

1.1.9 Overview of GMT 4.5.9 [Jan-1, 2013]

Predominantly a bug-fix release, we also have made some changes to GSHHS. First, GSHHS is now called GSHHG, the “Global Self-consistent Hierarchical High-resolution Geography”, since GSHHG contains both shorelines as well as political boundaries and rivers. GSHHG, being required by both GMT 4 and 5, is now released separately from GMT. Second, we have made some minor changes to a few islands that have shown to be offset with respect to modern data (Tahiti, Moorea, and Mehetia in South Pacific and Agalega Islands in the Indian Ocean). Third, we have removed the item known as “Sandy Island” in the Corel Sea since available satellite data show no evidence of land in this area. Finally, we have purged 48 duplicates of very small islands (mostly in the Red Sea, Persian Gulf, and the Cook-Austral archipelago) where inaccurate WDBII versions and accurate WVS versions of the same features had survived our initial processing. This GMT release thus coincides with the latest GSHHG release version 2.2.2. We also fixed a typo in the MJD date in the gmtdefaults.txt documentation and updated the CM4 coefficient files for the mgd77 supplement.

Below is the list of bug corrections for individual programs:

gmt_grdio.c
: For -R...r the adjusted w/e/s/n for a grid could have excessive e/n values, leading to extra work and large intermediate grids.
gmt_map.c
: With -R...r and -JE, pscoast could end up trying to paint a block that is outside the rectangular region yet whose projection completely contains the region. Clipping function could shift the w/e boundaries by an extra 360, thus finding nothing inside in some pscoast tiles.
gmt_shore.c
: Will now look for either “.nc” or “.cdf” GSHHG files; needed since GSHHG is now stand-alone. Also look in given dir directory if compiled in via configure with-gshhg-dir=dir.
gmt_support.c
: Improved contour orientation function for grdcontour -F, thanks to Thomas Hottendorff. Resolved case of crossings at perfect saddles.
gmtmath.c
: Operators UPPER and LOWER did not assign output to rows that contain NaN.
grdfilter.c
: For -D4 we must set scale based on max |lat| in input grid.
grdproject.c
: The north coordinate would get reset to 2 when -C was used, thus creating wrong map setup calls.
project.c
: Had to backport changes implemented in GMT 5 to make the -C-T-G combination work properly.
pslib.c
: Typo in array index in European letter shorthands for German double-s; thanks to David Lachapelle.
psxy.c
: Would hang for -Jx...d due to wrong meter-per-degree variable.
meca/psvelo.c
: Bug in parsing the -A arguments.
mgd77/mgd77.c
: Copied 4 rather than 2 characters into day field, causing overflow and SEGV.
mgd77/mgd77magerf.c
: The -A+ttime option did not parse ISO calendar items.
misc/dimfilter.c
: Initialized a max value to -DBL_MIN instead of -DBL_MAX.
spotter/backtracker.c
: First and last point in a path was not converted back from geocentric coordinates.
spotter/libspotter.c
: Did not check for time reversals among total reconstruction rotations.
spotter/originator.c
: Bad format statement would wreck the 5th output column when run in default mode.
x2sys/x2sys_cross.c
: Did not convert time to seconds before calculating speed. Disallowed the distance gap check if time was present.

1.1.10 Overview of GMT 4.5.8 [Apr-1, 2012]

Another bug-fix release, except for the mgd77 supplement where we now have added support for the new MGD77T tab-delimited format introduced by NGDC, and for ps2raster.c under Windows where the Ghostscript executable path is now fetched from the registry. In case that fails, we fall back to the old “get it from the path” mechanism. One common bug shared by several programs was the failure to consult FIELD_DELIMITER and/or D_FORMAT for ASCII output formatting. Below is the list of bug corrections for individual programs:

gmt_customio.c
: Change nx and ny to “unsigned short int” type in surfer6 header. Note that original format specification by Golden Software clearly say they are “short int” but this change shouldn’t break anything and will allow dealing with larger grid sizes.
gmt_grdio.c
: The opening of files for rec-by-rec grid reading had a mixed if-test that inadvertently could take us to the wrong else clause.
gmt_map.c
: -R0/360/... and -JMlon/lat would not set global_map properly, giving NaNs as scale.
gmt_plot.c
: Fixed the corner caps in linear basemap frames. In function GMT_xy_axis, string[GMT_CALSTRING_LENGTH] was used to hold labels that could be longer, leading to memory corruption.
gmt_proj.c
: For -JE, both projection center and its antipode gave x = y = 0. Now, antipode (which maps to a circle) results in NaN, NaN. Determine whether conic projections are north or south polar by looking at the selected region, not just the central point.
gmt_support.c
: Function GMT_crossover was susceptible to longitude wrapping. Function GMT_getfill got confused when a Windows path C:/etc was given as an pattern file (mixed up with :F and :B mechanism). In GMT_hold_contour_sub, a variable called closed (which could be a flag 0-3) was used as a 0/1 variable. Added variable is_closed = 0|1. Fixed error in reading colors of patterns in -G options. GMT_inonout_sphpol_count did not properly handle line segments that were exactly vertical in all cases, leading to errors as to a point being inside or outside a spherical polygon.
grdcut.c
: Now recognizes any region to be a global grid as long as nx*dx == 360. The -Z option would give incorrect region for pixel grids; print warning if output grid equals input grid (i.e., no change).
grdfilter.c
: Now recognizes any region to be a global grid as long as nx*dx == 360.
grdimage.c
: Now recognizes any region to be a global grid as long as nx*dx == 360.
grdinfo.c
: Now recognizes any region to be a global grid as long as nx*dx == 360.
grdlandmask.c
: Now recognizes any -R as a global grid as long as nx*dx == 360.
grdview.c
: Contouring could suffer from the same round-off issues that affected grdcontour (and fixed back in 2004). Now the same fix is applied here. Also let -S default to 0 as stated in man page; this matches the default in grdcontour. Also, when nodes are adjusted to avoid matching a contour value exactly (during contouring), the same adjustment must be made later when those nodes are used to determine how to stitch together polygons for fill.
minmax.c
: Avoid infinite loop if a record has different number of fields than expected.
nearneighbor.c
: Now recognizes any region to be a global grid as long as nx*dx == 360.
project.c
: For -Ginc -bo we failed to write any output.
xyz2grd.c
: On Windows it never actually opened the input data file, so a crash resulted down the road. Only require two input columns when using -An.
meca/pscoupe.c
: Could not read from stdin because of the cross dll boundaries on Windows, must use GMT_stdin instead of stdin.
meca/pspolar.c
: Could not read from stdin because of the cross dll boundaries on Windows, must use GMT_stdin instead of stdin.
meca/psvelo.c
: Did not show -A in the synopsis. Could not read from stdin because of the cross dll boundaries on Windows, must use GMT_stdin instead of stdin.
mgd77/mgd77.c
: Fix memory allocation bug in MGD77_Read_Header_Sequence where it was reading MGD77_RECORD_LENGTH records into a shorter MGD77_HEADER_LENGTH longer variable.
mgd77/mgd77list.c
: Confused the meaning of the -F shorthands mgd77 and all. Now handle all stored and derived quantities needed to reproduce original data files.
mgd77/mgd77track.c
: Did not set GMT’s time system to Unix before dealing with dates. Now done centrally by MGD77_Init.
x2sys/x2sys_cross.c
: A memset call used wrong size if 64-bit, thus not resetting some boolean values causing crossovers to be missed.

1.1.11 Overview of GMT 4.5.7 [Jul-15, 2011]

This is another bug-fix release, including an update to GSHHS (now at 2.2.0) which fixes a truncation error for the polygon areas (which only affected users of the gshhs supplement and not GMT itself). The supplement tool gshhs now has a few more options to allow better feature extraction for GSHHS users. Below is the list of bug corrections:

gmt_agc.c
: Did not do proper indexing for complex data. Had wrong size for array floatvalue.
gmt_grdio.c
: Failed to create proper old-style v3 netcdf file if selected for as output format in grdblend. Did not account for the doubling of a grid array for complex data when scaling data after read. This made grdfft give odd results when grids with a scale other than 1 was read. Bug was first introduced in GMT 3.0 in 1995, making it a bug with seniority!
gmt_map.c
: The GMT_wesn_search did not handle periodic longitudes well, now replaced with proper quadrant checking as in minmax. GMT_wesn_clip needed to adjust longitudes to fit the given domain. Clip path for van Grinten was wrong for global maps.
gmt_mgg_header2.c
: Did not implement complex arrays at all.
gmt_plot.c
: Force arcs to be clockwise in GMT_pie and GMT_matharc.
gmt_support.c
: Hardwired DEG_TO_KM with Earth’s mean radius meant wrong distance results for other planets. Now using the current ELLIPSOID values.
pslib.c
: Did not restore to current font size after a sub- or super-script if a font size change had previously taken place. Also did not recompute sub- or super-script sizes after a font size change. [Thanx to Christian Sperber for noticing].
grdcontour.c
: Could crash if -C -A give no contours and we freed a non-allocated array. [Thanx to Walter Harms].
grdblend.c
: Now mentions which formats are supported and polices the process.
grdedit.c
: The -N option did not consider replicate w/e points for gridline-registered grids.
grdfilter.c
: The -Nr option did not work as it skipped assigning the NaN. Convolution weights were y-transposed so gave wrong results at/near the poles for spherical filtering. Also, some nodes were duplicated in the convolution, resulting in inconsistent values at the S or N pole. Now gives consistent values at poles and along shared E/W cols.
grdimage.c
: When image is grayscale and -Q is used then image must be converted to 24-bit and we set NaN color to a non-gray value.
grdlandmask.c
: Did not correctly truncate nodes for GSHHS bins with no data.
grdproject.c
: Output header had wrong units when non-inch settings were in effect.
grdview.c
: Removed grdview_init_setup because it yields unpredictable results and prevents lineup of different 3D plots.
mapproject.c
: Kept treating x,y as lon,lat when -G[lon/lat/]c (c for Cartesian) were given.
nearneighbor.c
: The -N1 selection did not reset the minimum sector setting to 1. For global region in longitudes, cannot extend w/e limits as we do for Cartesian and region-limited areas.
project.c
: The -G-T combination failed to produce great circles.
pscontour.c
: Got wrong contour level when some segments along edges were skipped.
pstext.c
: Option -L was missing from the synopsis. The -A option resets the text angle but also needed to reset justification.
xyz2grd.c
: Fixed bug when reading a file using -E option on Windows.
triangulate.c
: Did not check that input file could be found before trying to compute stuff and crash [Thanks to Orion Poplawski].
spotter/grdrotater.c
: Ensure that output grid xmin/xmax honors the current OUTPUT_DEGREE_FORMAT range settings. Could end up with w=e=0 in some cases.

1.1.12 Overview of GMT 4.5.6 [Mar-1, 2011]

This is another bug-fix release, including an update to GSHHS which fixes error is the Germany-Poland border and a few “spiky” islands. Therefore, this version requires the new GSHHS 2.1.1 release. We also patched some errors in the “jet” color table. Below is the list of bug corrections:

gmt_init.c
: On Windows: Look for HOMEPATH after HOME for setting GMT_HOMEDIR. Processing of mathangle symbol (-Sm[bfl]) confused the unit detector. The symbol-parsing for psxy and psxyz did not properly set the column types when no symbol size was given. This affected symbols that require angles (-Sw, -SE, -SJ) when unit was SI. The -: option messed up the column type arrays; should only swap x,y data columns, not type columns.
gmt_cdf.c
: Apply netcdf fix for open/fill; thanks to Sebastian Heimann.
gmt_grdio.c
: Increase the size of the string array in GMT_grd_get_units to avoid “Buffer overrun” that occurred with long description strings.
gmt_io.c
: When ASCII mode, also need to save/restore any netcdf i/o settings.
gmt_map.c
: Fixed bug in Haversine equation for duplicate point. For van Grinten: errors in left/right-circle functions. We have a safety valve for preventing a painful slow search around the map perimeter. The search is appropriate for maps but not for mapproject results. The limit was for 200 inch wide maps == 14400p. A user ran pscoast with 14401p and was caught. Now check for 400 inch and also check current page size (PAPER_MEDIA). If map width vastly outsizes the paper size then it is probably a projection job.
gmt_nc.c
: Apply netcdf fix for open/fill; thanks to Sebastian Heimann.
gmt_plot.c
: A colored TICK_PEN would also color annotations. Tried to free an unallocated array in GMT_draw_custom_symbol.
gmt_support.c
: GMT_cspline should initialize c[0] = c[n-1] = 0.0 in case it is called repeatedly [this is not the case in GMT]. When calculating how far to place an annotation from the tick mark we must check if a fancy frame width exceeds the tick length. GMT_inonout_sphpol_count failed to detect crossings if a polygon had vertical line-segments with same longitude as the point we were testing.
pslib.c
: Add check for incomplete escape sequences..
gmtselect.c
: Due to resampling of parallels in -N, some points exactly on a coast bin parallel could fail the test due to roundoff. Fixed by not resampling coastlines since it is a Cartesian test.
grd2xyz.c
: Extended -Ef mode to write floats (patch by Pierre Cazenave).
grdclip.c
: Now complains if no -S option was given.
grdfilter.c
: Allow -D2 and -D3 to handle periodic and polar boundary conditions.
grdlandmask.c
: Same as entry for gmtselect.c.
grdmask.c
: Must skip “polygons” with less than 3 points. Also, the resampling distance for spherical data was wrong, now 0.1 degrees.
grdproject.c
: Not providing -R was not working anymore. Test if hemisphere sign is provided when doing -Ju and no -R. Now assign proper x/y units.
grdtrack.c
: Message about -L being obsolete should only come when no modifier is given; else it is valid for BC setting.
ps2raster.c
: Bug fix after last -F option update. Must pass the optional -C args when calculating BB. Added -dSAFER as well, + fix -F option [F. Wobbe]. For some reason -W was not forcing -A. Now it does it again.
pscontour.c
: Did not find some contours following triangle edges.
pslegend.c
: Need to keep the original -R-J around for proper calls.
psscale.c
: Now patterns have constant orientation regardless of using horizontal or vertical bar.
nearneighbor.c
: Changed default to a more reasonable -N4/2.
splitxyz.c
: Did not list -Q in the synopsis.
surface.c
: Bug if using seconds (c) in search radius (got minutes).
triangulate.c
: Did not use projected coordinates when -R-J was given.
xyz2grd.c
: Do not tolerate NaNs in x,y and give error (e.g., if junk is given). Failed if -Evalue was given and the ESRI grid already had a nodata-value line. Now will process this line, if present. The value given on the command line will override any setting found in the file. Also made string-checks case-insensitive.
meca/utilmeca.c
: Patch to fix incorrect plotting of moment tensors with big isotropic components. Thanks to Jeremy Pesicek. Fix bug affecting the plot of P and T axis.
mex/grdwrite.c
: Round-off could lead to false detection of a non-equally spaced grid.
misc/gmt2kml.c
: Options -N+ and -D would crash under Windows (usual DLL hell).
spotter/backtracker.c
: For -W, now report full-length major/minor axes and not SEMI-axes (docs said major/minor but code did semi.)
spotter/grdrotater.c
: Did not handle the rotation of an entire global grid since the polygon outline interfered with the domain.
spotter/rotconverter.c
: Forgot to skip args when -N or -S was used.
x2sys/x2sys.c
: Minor bug in x2sys binary reading of floats.
x2sys/x2sys_datalist.c
: When geographic data and -R it failed to consider periodic longitudes.
x2sys/x2sys_get.c
: The -N option did not work properly, and the reported Y|N flags reflected the entire track on not just the portion inside the region. Man page updated to clarify what is returned.
x2sys/x2sys_init.c
: Did not write both of -Nd-Ns to the tag file. Crashed if -D was not given [should imply -DTAG].

1.1.13 Overview of GMT 4.5.5 [Nov-1, 2010]

This is again mostly a bug-fix release, and coincides with the availability of GMT 5.0.0α. Due to a few issues we had an aborted update to 4.5.4 that was never announced; hence the 4.5.5-numbered version. A few minor improvements have been added:

  1. The spotter supplement now converts geodetic latitudes to geocentric before doing spherical rotations and recovers geodetic coordinates for output (this new behavior can be bypassed by setting ELLIPSOID to Sphere). Thanks to L. M. Matias for pointing this out.
  2. We have added time-axes support for the Hawaiian language (Thanks to Kāwika Trang).

Here is the list of bug corrections:

configure
: Now sets correct mex extensions for 64-bit operating systems.
gmt_init.c
: The -B labels would not tolerate use of the text escape sequence @: (for changing font size).
gmt_map.c
: Did not check if nodes were beyond the horizon in GMT_grd_project. Also did not initialize output grid to NaNs before filling.
gmt_plot.c
: Bug in fault symbol psxy -Sfrc fixed (thanks to J. Robert). Also, GMT_map_latline and GMT_map_lonline functions tried to draw two-point lines when in fact no points were defined.
gmtdefaults.c
: The -D options would crash under Windows.
grdmask.c
: The -Ap|m options were ignored since the mode was not checked.
grdpaste.c
: Lacked -fg so could not paste 352/360 and 0/8 in longitudes.
grdtrack.c
: Did not ensure that given -R was adjusted to fit grid spacing.
mapproject.c
: Did not show/explain the option of appending + to -L. Corrected synopsis, usage, and man page. Did not reset azimuth to NaN at start of new segment.
minmax.c
: With -I, could end up returning -R355/0/... since 360 became 0.
nearneighbor.c
: Did not check if -S had not been set.
psbasemap.c
: The -L option had trouble parsing if there were + signs within the label string.
pscoast.c
: The -L option had trouble parsing if there were + signs within the label string.
ps2raster.c
: Made tolerant of \r-only line-endings which caused trouble before. The -A- option did not reset -A for -W.
psimage.c
: The justify text variable must be 3-char longs to hold trailing 0. This caused SEGV on some systems.
psmask.c
: Did not warn if clipping levels were not restored in last overlay.
pstext.c
: Added missing description of -A option.
.html]psxy[z].c
: Units given in -S without sizes (e.g., -Sci) would be ignored and overridden by MEASURE_UNIT. The -Ap|m options were ignored since the mode was not checked.
mgd77/mgd77manage.c
: The -F option had no break statement to prevent fall-through.
mgd77/mgd77track.c
: Had inactive code to write segment header to output.
mgd77/mgd77list.c
: The -G option was not listed in synopsis or usage, only in the man pages. Also -Fall+ and -Fmgd77+ did not append the auxiliary columns properly.
mgd77/mgd77magref.c
: The -D option failed on numeric arguments.
misc/gmtstitch.c
: Could crash if -C was used.
misc/gmt2kml.c
: Did not parse -D“description” for points. Only append running number when a segment has more than one point, else just use segment label.
spotter/rotconverter.c
: Complained of “bad option” when a rotation with a negative longitude was given on the command line, e.g., -135/35/-2.5. Would sometimes issue a rotation twice (for the same time).
sph/Makefile
: Did not have LDFLAGS in link statement.

Also, Appendix F had missing shading for two items in the Standard+ table, and example 23 placed the city names at an angle of 1 degree rather than horizontally.

1.1.14 Overview of GMT 4.5.4 [Nov-1, 2010]

A few minor technical issues in the distribution led us to make a few changes and increment the version to 4.5.5.

1.1.15 Overview of GMT 4.5.3 [Jul-15, 2010]

This is mostly a bug-fix release, including more corrections to the political boundaries distributed via the GSHHS netCDF files (these affect the Syria-Israel, Israel-Jordan, Moldova-Ukraine, and the Eritrea-Ethiopia borders) as well as missing river-lake metadata in the GSHHS distribution. Therefore, this version requires the new GSHHS 2.1.0 release.

Here is the list of bug corrections:

configure
: Fixed reversed use off –enable-flock.
gmt_init.c
: Chop off any eventual EOLs characters that might be in argv strings as it will happen when it was created by a shell command. We need this so that native Windows binaries can be used in Cygwin.
gmt_io.c
: GMT_is_a_blank_line saw “t” instead of TAB as whitespace. Added GMT_io.skip_duplicates [FALSE] to control if consecutive records with identical x,y should be skipped. This is needed by programs that uses GMT_sph_inonout, which does not expect to find duplicates vertices. GMT_fgets now checks for input record truncation and handles this gracefully (gives warning and winds to next record).
gmt_map.c
: Tried to free memory when nothing had been allocated. GMT_wesn_clip function would clip polygons even though there were no restrictions on longitudes (w/e = 360).
gmt_plot.c
: Parallels that should be straight (e.g., in -JI) would sometimes appear with jump gaps. Fixed bug in GMT_plot_map_scale that could lead to endless loops when using scales to 100 km or any exact power of 10. Error was limited to 64-bit.
gmt2rgb
: Option -G was freeing the output name before it was even allocated.
grdcontour.c
: The L or H color for first min/max annotation was not set. Placement of H and L annotations improved by using centroids.
grdmask.c
: Did not handle periodic longitude input when -fg was used.
grdview.c
: Fixed bug in parsing of -W[m|c|f] option when color starts with [m|c|f]. Check that topo and illumination file have the same size, otherwise it would crash.
greenspline.c
: Must insist that one of [-R-I], -N, or -T is specified.
mapproject.c:
: Applied scaling to -Cdx/dy when -Fk was used, despite docs saying -C is in meters when -F is used. Fixed, and clarified docs/man to say with -F, -C is always in meters.
nearneighbor.c:
: Did not handle periodic longitude input when -fg was used.
ps2raster.c:
: Now checks that all PS files begin with %!PS. End matter did not get parsed when there is no %%Orientation.
pshistogram.c:
: Fixed incorrect bin count when a datapoint equaled xmax.
pslegend.c:
: Uninitialized text string could put garbage in script.
psmask.c:
: Did not handle periodic longitude input when -fg was used.
psxy.c:
: For -Svs, the 2nd set of coordinates did not obey -:. The -Sw|W symbols did not handle the azimuth/direction conversions properly. Added better handling of dimensions with units passed via columns in the data file.
psxyz.c:
: For -Svs, the 2nd set of coordinates did not obey -:. The -Sw|W symbols did not handle the azimuth/direction conversions properly. Added better handling of dimensions with units passed via columns in the data file.
surface.c:
: Did not handle periodic longitude input when -fg was used.
meca/psmeca.c
: Removed out of place and repeated line to compute size in -a option.
meca/submeca.c
: Replaced calls to d_atan2 by d_atan2d since the code expects angles in degrees.
mgd77/mgd77.c
: Incorrectly added track list =tracks.lis as another track name after correctly including all the listed tracks. No harm done other than an annoying “Cannot find track =tracks” message.
mgd77/mgd77magref.c
: Fix bug in -A option when using const time in calendar format.
mgg/mgd77togmt.c
: Now has proper synopsis.
misc/kml2gmt.c
: Did not anticipate optional attributes for tags like <PlaceMark >, etc.
sph/sphtriangulate.c
: Incorrect items for cols 3–4 for -N.
x2sys/x2sys.c
: Need to include the “.” when checking if a suffix is present in a filename. Reading of data formats .gmt and custom returned all columns and not just the requested columns, causing errors upstream.
x2sys/x2sys_datalist.c
: Check to see if both lon and lat had been requested only checked for longitude (twice).
x2sys/x2sys_list.c
: Implemented -S[+] to print info relevant to both cruises.

Here is a list of the recent enhancement to various programs; these were introduced to correct mistakes or overcome limitations:

  1. gmtmath.c has added function SQR (square).
  2. grdgradient.c now lets -S work alone without requiring -G.
  3. grdmath.c] has added function SQR (square).
  4. pswiggle.c -Dxgap now allows gaps to be in projected distances.
  5. mgd77/mgd77.c was updated for 11th generation IGRF – IGRF2010.
  6. x2sys/x2sys_get.c needed –L+[list] so internal crossovers can be added.
  7. GMT_nighttime.cpt color table donated by Andreas Trawoeger.
  8. GMT_paired.cpt qualitative color table by Cynthia Brewer.

1.1.16 Overview of GMT 4.5.2 [Jan-15, 2010]

This is mostly another bug-fix release, including one that required us to add more meta-data to the GSHHS coastline netCDF files. Therefore, this version requires GSHHS 2.0.2 or higher. As was the case for 4.5.1, note that the GSHHS polygons themselves have not changed (still at version 2.0). We also added in the relatively recent Nunavut province boundary in Canada. However, some enhancements were added as well, most notably a new graph frame mode for linear projections (to add arrow heads to math axes) and a new symbol in psxy.c (to draw a circular arrow used to indicate angles); these capabilities are demonstrated in a new (and final) example 30. Finally, we fixed the long-standing problem of psxy -SE requesting major and minor axes but actually treating them as if they are semi-axes. We now consistently expect and use major and minor axes; you may thus notice a scaling of two if you continue to give semi-major/minor axes. Here is the list of bug corrections:

configure
: Fixed bug with –rpath.
gmt_customio.c
: Fixed bug resulting from releasing the pointer to from_gdalread structure before its members were freed.
gmt_gdalread.c
: Force computation of min/max since metadata info may be wrong.
gmt_grdio.c
: GMT_read_img did not apply swab if little-endian architecture.
gmt_io.c
: GMT_access did not check for NULL filename.
gmt_init.c
: Now guards against getting a negative hash value, which happened when text was Russian language codes from ru.d. If GMT_DATADIR was set to a list of colon-separated dirs then init failed since we tried to check access as if GMT_DATADIR was always a single entity (as in the past). Opened file with fopen but closed with GMT_fclose. Bug in the parsing of -Jglon/lat/radius/lat. Both .gmtdefaults4 and .gmtcommands4 were assumed to be in UNIX format. Now we properly chop off Windows or Unix end-of-line characters.
gmt_mgg_header2
: Opened file with GMT_fopen but closed with fclose.
gmt_plot.c
: qsort of GMT_LONGs was passed int arrays. The calculation of the actual plot width of a map scale did not account for the effect of a 3-D view angle.
gmt_shore.c
: Changes to accommodate new GSHHS2.0.2 netCDF files which needed more metadata to properly compute the level of tile corners after features where dropped due to size, etc.
gmt_support.c
: Increase the size of the variable that contains the path to a CPT file and the pathname to BUFSIZ bytes.
gmtconvert.c
: The -S option failed for actual matches; the 2009/5/26 change screwed that up. Now fixed and tested.
grdcontour.c
: The -Q option should only apply to closed contours, and -T failed to find an inside point for some oblique projections.
grdgradient.c
: Now, -Lg will imply -fg to properly set geographic units. Fixed bug where the gradient at the south pole was not replicated to x = east.
grdlandmask.c
: Round-off and bug caused missing nodes for -F with -Rd.
grdtrack.c
: The -Z option gave z-values a longitude formatting, including 360-degree wrapping.
mapproject.c
: The -S option failed after recent i/o makeover.
pslib.c
: Ellipse was wrongly dimensioned by semi-major and semi-minor axes, instead of major and minor axes. Also, memory never got freed by ps_free.
psrose.c
: Needed to set scale to 1 so the bounding box calculations would be correct for EPS output.
psscale.c
: The -Bg now correctly produces gridlines using GRID_PEN_PRIMARY.
psxy.c
: Incorrectly drew tips at plot boundary when clipping the error bars.
psxyz.c
: Fixed bug in distance sorting (also made much simpler).
imgsrc/img2google
: One gmtmath call did not have the required -Q.
mgd77/mgd77sniffer.c
: SEGV when an array index was allowed to become -1.
misc/gmt2kml.c
: Removed rogue newline before writing coordinates and an additional > in a tag.
misc/gmtstitch.c
: NULL segment headers were passed to strcpy() to give SEGV; also fixed output message when -C was used and only one segment was present.
misc/kml2gmt.c
: Mix of fopen and GMT_fclose not good under Windoze.
sph/sphinterpolate.c
: Bad index on line 317 went outside array limit. Also did not initialize the grid header properly.
sph/sphdistance.c
: Did not initialize the grid header properly.
spotter/backtracker.c
: When -L is set, -mo must be turned on automatically even if -m is not set.
spotter/grdrotater.c
: Bug in node index resulted in no longitudinal variation in the rotated grid.
x2sys/x2sys_init.c
: Did not accept -m, only the now obsolete -M.
x2sys/x2sys_datalist.c
: Did not retrieve the correct data columns when -F was used.
x2sys/x2sys_list.c
: Weights were not written out when names were selected as well, as in -Fncw.

Here is a list of the recent enhancement to various programs:

  1. fitcircle.c will let -S optionally take a fixed latitude instead of finding the best-fitting latitude [Default].
  2. gmtdefaults.c has new BASEMAP_TYPE = graph option for linear projections that wish to have their axes extended 7.5%, ending in arrow heads.
  3. grd2xyz.c has a new option -N which can be used to replace NaNs with another value on output.
  4. psxy.c now has a new symbol -Sm option for math angle, which lets user draw a circular arc with optional curved arrow heads at neither, one, or both ends.
  5. psxyz.c also has the new symbol -Sm option but does not draw the curved arrow heads yet.
  6. mgd77/mgd77manage.c adds attribute array of TZ corrections for the few cruises that stored TZ and local time instead of UTC time.
  7. spotter/rotconverter.c has new option -E to reduce stage pole opening angles by fact [0.5] (e.g., to get half-spreading rates).
  8. x2sys/x2sys.c now lets x2sys programs automatically strip off extensions for tracks given via list files. Also, various lists can contain header records.
  9. x2sys/x2sys_get.c has new option -E to select extensions on output, and -Q[i|e] to be used with -L. Finally, -D now only lists track names.
  10. x2sys/x2sys_solve.c has new option -Z to remove smallest dist/time for -Ed|t.

1.1.17 Overview of GMT 4.5.1 [Sept-20, 2009]

This is almost entirely a bug-fix release where we address several 64-bit incompatibilities and rebuild the netcdf GSHHS library to include some attributes from GSHHS that were needed by new options in pscoast and other programs. Note that the GSHHS polygons have not changed (still at version 2.0), but we had to update the derived netcdf repackaging used by GMT to 2.0.1. However, some enhancements were added as well, most significantly support for the polyconic projection (-JPoly), experimental support for grid and image imports via GDAL (requires –enable-gdal during configure and properly installed GDAL libraries and include files), and allowing -JXwidth/height to recompute a height given as zero based on the width (or vice versa) and the aspect ratio of the region.

Here is the list of bug corrections:

configure.ac
: Now use –enable-flock to enable file locking, instead of –disable-flock to disable if. Use –enable-64 to force 64-bit compilation, use –disable-64 to force 32-bit compilation; otherwise use default. Switch -DGMT_QSORT is now on only for OS X prior to Snow Leopard as the latter has a correctly working 64-bit qsort function.
gmt_customio.c
: Made surfer grid header i/o 64-bit compliant.
gmt_map.c
: Changed GMT_truncate from being a pointer function (PFL) to a regular function that simply calls GMT_truncate_x or GMT_truncate_tm depending on whether projection is TM. This to avoid problems with calling GMT_truncate with constant argument -1 in 64-bit machines and thus producing wrapped pscoast maps.
gmt_shore.c
: The -A+r|+l option was premature as more info in the GSHHS netcdf files were needed to properly skip features. Now working with the revision GSHHS 2.0.1.
gmt_support.c
: The routine that checks if a points is inside or outside a polar cap had trouble when a point’s longitude exactly equalled one of the polygon points since round-off could cause our test to fail. Fixed bug in x_inc variable as function of latitude. GMT_log_array recoded to avoid 64-bit error (last value was not included).
grdcontour.c
: The -L option was not used in limiting contours.
minmax.c
: The -EH option did not work.
psmask.c
: With -D, the internal di, dj constants were set way too large.
pswiggle.c
: With -Jxscaled, distances were not scaled correctly so -D failed.
mgd77/mgd77magref.c
: Got wrong time as time initialization had changed but not implemented in this program.
mgd77/mgd77sniffer.c
: Bug in -I option; needed to set bitpattern for bad faa.
misc/gmt2kml.c
: Option -T did not handle spaces in titles and folder names. Under Windows, the output got scrambled due to DLL hell.
spotter/backtracker.c
: Failed in 64-bit mode due to variable mismatches.
x2sys/x2sys_get.c
: The -L option failed if list contained file extensions.
x2sys/x2sys_solve.c
: Did not allocate array for storing weights.

Here is a list of the recent enhancement to various programs:

  1. gmt_customio.c now has optional GDAL read-only interface as format 22 (code gd).
  2. gmtset.c has new default TRANSPARENCY = stroke/fill that can change the PDF transparency for stroked and filled items [Note: Only supported by Adobe Distiller].
  3. pslib.c has added option to reduce all color to gray scale via PS_COLOR = gray.
  4. grdimage.c has optional support that allow single-band image import via GDAL.
  5. misc/nc2xy.c now has -bo option.
  6. mgd77/mgd77magref.c can combine IGRF and CM4 computations.

1.1.18 Overview of GMT 4.5.0 [July-15, 2009]

This is another significant update of the official distribution and hence it has a mix of bug fixes and program enhancements. We have added a new supplement (sph) which offers interpolation, triangulation (Delaunay and Voronoi), and distance calculations on a spherical surface. The hard work is done by the original effort of Robert Renka who developed the Fortran-77 SSRFPACK and STRIPACK libraries; these are here supplied via a f2c-assisted translation. The imgsrc supplement has a new Bourne script img2google, which simplifies making Google Earth tiles from Sandwell and Smith bathymetry. The mgd77 supplement has a new program mgd77magref, which is used to evaluate either the CM4 comprehensive geomagnetic model, a more sophisticated alternative to IGRF, or the IGRF. The misc supplement has received two new tools (gmt2kml and kml2gmt) that simplify the presentation of GMT data in Google Earth, and one (dimfilter) that offers directional spatial filtering of grids. The x2sys supplement has a new tool (x2sys_merge) to merge updated COEs table into a main COE table database. Finally, ps2raster.c has evolved further and can now be used to create simple KML files for Google Earth.

A major new enhancement is the global option -g, which is used to determine if excessive spacing between data points (“gaps”, to be defined in a variety of ways) should be used to segment an otherwise continuous line. We expect to enable -g in several programs during the next revision; at the moment it is available in gmtconvert, mapproject, psxy and psxyz. Given that all the lower-case GMT options deal with low-level data i/o settings we have decided to rename the -M option (which controls the presence of multiple segment headers) to -m; this allows us to promote this ubiquitous option to global status (i.e., has the same meaning in all GMT programs). Use of -M will remain valid for the rest of GMT 4.x but results in a warning about the new usage. Related to this is the introduction of a new parameter (NAN_RECORDS) that determines if NaNs in key columns (such as longitude, latitude) should constitute a line break or bad data to be skipped.

We have revised how ellipsoids are specified. When importing an ellipsoid file, we allow a,b,f as ellipsoid parameters, where b or f could both be zero. If file does not exist, attempt to read name as a[/[b=|f=]f], meaning semi-major axis, b=semi-minor axis, f = flattening, or inverse flattening. We have also added parameters for the TOPEX ellipsoid and for the Moon and planets (IAU2000).

This release of GMT coincides with the release 2.0 of GSHHS, the coastline data used by GMT. In addition to general improvements to the data, we have expanded the -A option that controls the limits on what features to extract. New modifiers allow users to exclude “river-lakes” and any feature whose area is less than a fraction of the original full resolution feature.

Finally, our configure script continues to evolve and now better supports installation on 64-bit systems and can automatically detect if and where netCDF exists on your system.

Here is the list of bug corrections:

gmt_customio.c
: In GMT_ras_read_grd_info, wrong size was passed to fread.
gmt_grdio.c
: Fixed bug in GMT_grd_setregion: longitudes were limited inward instead of outward. Caused white-space left and right in grdimage.
gmt_init.c
: No longer remove supposedly empty .gmtcommands4 file. Could have been written to by piped GMT command. Now create/write only when new matter is to be written.
gmt_io.c
: When only -mi is set there should be no multisegment headers on output. yet there was no if-test to check for that.
gmt_map.c
: Fixed serious bug in GMT_az_backaz_flatearth: incorrectly converted degrees to radians. Fixed bug in GMT_get_rotate_pole: called GMT_rotate_pole_forward with radians instead of degrees. Incorrectly attempted to free array only used when antialias is on. Fixed inability to determine projection pole when central meridian and western boundary are the same (Albers, Conic equidistant). Incorrectly used central longitude to set default midpoint for 3-D maps using the -E...+ mechanism when data were not geographic. Numerous bug fixes related to UTM: Did not report bad zone for < A or > Z; Zones A+B reached till 84S instead of 80S; Zone J wrongly produced value for Zone K; Zones U+W were not recognized; Now also equates Zone O to Zone P (already equated Zone I to J).
gmt_mgg_header2.c
: There was no support for (a) floating point grids and (b) swapping happened based on machine byte order and not based on actual file byte order.
gmt_nc.c
: Argument [layer] was not stripped from varname and would end up in output grid. Set zmin/zmax to NaN when info not in header.
gmt_plot.c
: Crashed when using time labels in 3D plot.
gmt_proj.c
: Fixed bug in longitude computation when standard parallel is on Southern Hemisphere.
gmt_support.c
: Exceeded array bounds in GMT_read_cpt for hsv conversion. The 3-D view -E option used in many programs (such as psxyz) could not handle exponential notations. Determining of a color scale is continuous should be based on HSV (not RGB) values when read as HSV. The modifier :radius[unit] to the -G option in grdcontour for the placement of contour labels is now +rradius[unit] since the colon interfered with ddd:mm:ss coordinates. Given -I601+/601+ the y-increment was not processed correctly. Fixed bug when x_inc in km, m, etc, the value was ignored (always 1).
blockmean.c
: Gave bogus usage about number of input columns when -E is set but -E only affects output. Same for blockmedian.c and blockmode.c.
gmtconvert.c
: With -S, did not report the number of output segments.
gmtmath.c
: Tried to free memory that was not allocated.
grdfilter.c
: Toggle -T was not processed correctly if -Rgridfile was given.
grdgradient.c
: For geographic grids, make sure N and S pole only has a single value on output.
greenspline.c
: Calculation of nz layers was off by one, leaving out the penultimate layer.
grdimage.c
: We now make sure the NaN color is unique when -Q is used.
grdmath.c
: The PDIST operator did not handle binary input files (as LDIST does).
ps2raster.c
: Exceeded memory allocation for out_file. Now in static memory.
psclip.c
: Produce at least “S V” when the clippath is empty.
pscoast.c
: Produce at least “S V” when the clippath is empty.
psscale.c
: When using -A with a vertical color bar, the annotations where left- in stead of right-aligned. When using -A on horizontal bar, the last tick mark would not appear.
gshhs/gshhs.c
: Checked wrong variable to test if a file was given on the command line.
meca/util_meca.c
: Needed to convert the output from d_atan2 to degrees.
mgd77/mgd77.c
: Was failing to read Windows terminated ASCII files. Fixed bug in MGD77_igrf10syn routine (mistake done while cleaning the f2c version) that resulted in altitude always being on Earth surface when geocentric coordinates were used. Do not use any GMT i/o-functions to access ascii files on creation/writing so that it can work under Windows as well. Converting files from netCDF to MGD77 where time = NaN gave junk records.
mgd77/mgd77list.c
: Used GMT_LONG in memset for an array allocated as int. Used wrong time for IGRF calculation; worst case situation was off by 1 year.
mgd77/mgd77manage.c
: The -D option would remove original attributes. Now both original and revised attributes are carried in memory. This problem also affected mgd77convert -FC if used on a file after mgd77manage -D had been run. Used wrong time for IGRF calculation; worst case situation was off by 1 year. Now uses GMT_read_img to handle *.img files; earlier the inline code would fail on the new extended img files.
mgd77/mgd77sniffer.c
: Did not reset the E77 structure between cruises.
misc/gmtstitch.c
: Did not read standard input if no files were given. Now honors -V properly.
x2sys/x2sys.c
: Wrong array argument passed for correction aux values.
x2sys/x2sys_cross.c
: Removed the -F option which was not honored anyway.
x2sys/x2sys_init.c
: Now properly copies the *.def file to the TAG dir.
x2sys/x2sys_datalist.c
: Only ASCII output had corrections (-L) applied.
x2sys/x2sys_put.c
: Fixed -D option (was falling in a endless loop).

Here is a list of the recent enhancement to various programs:

  1. gmt_init.c has added support for enhanced -E option for 3-D perspective views which allows the specification of a fixed point (needed for new Default setting NAN_RECORDS = skip|pass [skip]. If “pass” we treat let the programs handle the NaNs; for some programs the NaNs in input record will act as indicators of data gaps for continuous lines. If “skip” we report them as bad records. In both cases the records are skipped. Allow negative integer interval (-n) for annotating log axis; this means annotate every n’th power of 10.
  2. gmt_nc.c will properly handle netCDF that have LatLon = 0, 1: flip x and y.
  3. gmt_shore.c Added support for the +ppercent modifier to limit features whose area is less percent of the corresponding full-resolution polygon.
  4. gmt_support.c enables GMT_intpol to handle NaNs by treating them as segment boundaries. Rely on Shewchuk’s triangle function to get Voronoi output.
  5. blockmedian.c now returns the quantile(x), quantile(y) location when -T is used. Added -Eb to get box-and-whisker output (0. 25, 50, 75, and 100% quantiles).
  6. gmtconvert.c now allows -S~pattern which reports segments whose header does NOT contain pattern. Should pattern actually start with ~ we escape it with \~pattern.
  7. gmtselect.c can now handle dateTclock strings in -Z when used with data whose 3rd column contains time.
  8. gmtmath.c added MOD function (remainder after floored division, Knuth style). This is in contrast to FMOD which gives the remainder after (horror) truncated division.
  9. gmtselect.c Same upgrade to -A option as pscoast.
  10. grdcut.c has new option -Z which is used to determine a rectangular subregion so that the rejected area have values entirely outside the given z-range.
  11. grdfft.c now also has a Butterworth band-pass filter.
  12. grdinfo.c will now use plain text (and not code) to report the file format used. Add option -L0 to actually scan data to determine zmin and zmax.
  13. grdlandmask.c Same upgrade to -A option as pscoast.
  14. grdmath.c added MOD function (remainder after floored division).
  15. ps2raster.c takes -W+k to create a simple KML file for Google Earth. Several other modifiers help to populate the KML file.
  16. psclip.c has a new option -T that turns on map region clipping without any input data files; it is a shorthand for -N/dev/null.
  17. pscoast.c has enhanced -C[l|r/]it fill offers to paint river-lakes separately from lakes. The -A option can be used to exclude river-lakes or lakes for level 2. -W now allows different pens for the 4 levels of shorelines.
  18. pglegend.c allows S record not to have text. Until now, when left empty would use whatever text was set to in previous record. The > record no longer needed before T; T, L, S, H alignment harmonized; I, M, B spacing improved; -B option added.
  19. psxy.c added -g to break lines into segments based on gap criteria; added -T to be a shorthand for reading no input, where we used /dev/null before.
  20. psxyz.c added -g to break lines into segments based on gap criteria.
  21. triangulate.c has new option -Q to generate Voronoi polygon edges.
  22. mgd77/mgd77convert.c allows users to give file.ext on command line.
  23. mgd77/mgd77list.c can now take -DA and -DB which, in the presence of time = NaN, will not output such records [-Da|b will]. Also clarified the ranges implied by -D and -S. Added -Fytime which gives decimal absolute year for time output. Modifier -At added which attempts to create fake cruise times based on header information and distance along track.
  24. mgd77/mgd77manage.c added -AE to ignore the verification status and process e77 anyway.
  25. mgd77/mgd77path.c added -P- to just list the IDs and not the full paths.
  26. mgd77/mgd77track.c added -Gt|dgap to recognize gaps in tracks based on distance or time between successive points.
  27. mgg/mgd77togmt.c added new -T and -W options to store total field and account for the magnetometer tow distance. Also tries to get info from header file.
  28. misc/gmtstitch.c has enhanced -T option to eliminate connections when the 2nd closest pair is too close. Also added -C option to simply separate the open from the closed polygons (no stitching). Added -L option to write out segment-link information.
  29. x2sys/x2sys_list.c can now accept list of weights for each track and output the composite weight for each crossover.
  30. x2sys/x2sys_solve.c has option -W means an extra column with crossover weight.

1.1.19 Overview of GMT 4.4.0 [Feb-15, 2009]

This is a significant update of the official distribution and hence has a mix of bug fixes and program enhancements. We have added a new program (greenspline.c) which offers interpolation and gridding in 1–3 dimensions using Green’s functions of various splines. Also, the misc supplement has a new tool (gmtdp.c) which offers line-reduction using the Douglas-Peucker algorithm we used for the various shoreline resolutions. The mex supplement has a new Matlab/Octave function (imgread.m) to directly read Sandwell/Smith *.img files. The x2sys supplement has three new programs: x2sys_list.c can extract a subset of crossovers from the list produced by x2sys_cross.c, x2sys_report.c reports statistics of crossovers, whereas x2sys_solve.c will determine systematic trends from a set of crossover errors. These programs are intended to replace the old x_system tools x_list.c, x_report.c and x_solve_dc_drift.c. We have also temporarily added GMT_qsort which is a 64-bit compliant version of qsort. The latter is broken under OS X 64-bit and is thus substituted on that platform only for 64-bit compilations until Apple fixes the problem. Finally ps2raster.c can now be used to create geotiff images if gdal is installed on your system. Here is the list of bug corrections:

gmt_customio.c
: Fixed sub-region access in Surfer format. This bug would manifest itself mainly when doing a grdcut with a N-S sub-region.
gmt_init.c
: Modified special checks for FreeBSD by also considering _AMD64_. -JXh was misinterpreted whereas -JXv was OK.
gmt_io.c
: Did not properly apply PLOT_DEGREE_FORMAT=ddd.x for decimal degrees.
gmt_map.c
: Determining where parallels and meridians intersected the map boundary was susceptible to roundoff for very small regions. Added improved clipping for geographic polygons using the Sutherland and Hodgman algorithm when there are no map jumps in longitude. Fixes problem with tiny strips of “land” along map perimeter for some projections.
gmt_math.h
: Check for macro definitions for system math functions.
gmt_plot.c
: Wrongly checked for map jumps for non-periodic map boundaries. Bug in GMT_fill_polygon that affected polygon outline. 3D text box was computed incorrectly. Could exceeded array size in GMT_epsinfo.
gmt_proj.c
: Fixed bug in Lambert conformal conic projection for southern hemisphere.
gmt_support.c
: Function GMT_get_arc did not check for division by zero. The -Gxfile:radius[unit] option in grdcontour.c passed the entire argument as the file name. Function GMT_polygon_is_open did not test for empty polygons (n = 0). Avoid interpolating hue (converted from RGB) over more than 180 degrees. Changed GMT_rgb_to_hsv to integer logic to avoid errors on some compilers. As a result: much shorter code as well.
gmt_vector.c
: Function GMT_resample_path would add 360 to points along meridians.
pslib.c
: ps_polygon can only split line when rgb[0] == -1 not < 0 since -3 now means to use a fill pattern. Redefined PostScriptcode for circle which needed a stroke (S) first, otherwise a line would be drawn from the previous symbol. Updated PSL_prologue.ps version. Image placement now in integers. Ensures that placement is consistent with e.g. box drawn with the same coordinates. Does not produce colormap with number of pixels and colors is the same.
gmtmath.c
: The D2DT2 operator whose boundary condition yields 0 should yield NaN if one or more of the nearby nodes are NaN.
grdblend.c
: A side-effect of the 2007-02-01 fix was that when the file is re-opened the row range is reset. Now the possible offset is computed during initialization but applied when the file is finally opened for reading. Better treatment of longitude periodicity if -fg is selected. E.g., if -Rg is used and a grid is -30/30 in longitude the output grid will consider 0-30 and 330-360 correctly
grdfilter.c
: -D5 did not initialize xscale so filter search box was set to region width which typically is much larger. The bug did not affect the results but unnecessarily increased runtime. Complained if -R was used and xmin was less than grid xmin for a full 360-range grid.
grdimage.c
: Failed to determine boundary of projected grid with enough precision.
grdmath.c
: Some 2nd-order derivatives whose boundary condition yields 0 should yield NaN if one or more of the nearby nodes are NaN.
grdreformat.c
: Since there is no longer a share/conf/gmt_formats.conf the usage message crashed. Did not initialize grid header structure and could get netCDF error “Named variable does not exist in file”. This could also occur in grd2cpt.c, grdedit.c, grdtrack.c, grdvector.c, grdview.c, and grdvolume.c.
psbasemap.c
: The syntax for the -L option had to change since one could not easily use the :label: specification if the coordinates were given in dd:mm[:ss] format. A new syntax has been implemented where one or more +?[args] strings are appended after the required parameters (see man page). Erroneously suggested that 3-D base was plotted at z=0, instead of at the bottom end of the z-axis.
pscoast.c
: The -Q option incorrectly required -J. Also, see revised -L as for psbasemap.c.
pslegend.c
: Now use Unix remove function to delete script after completion; this avoids a Windows problem.
psmask.c
: Fixed a bug for -D-file which did not write multi-segment headers.
psscale.c
: Logarithmic scale did not function properly when scale bar was vertical. When -I and -Li was used we did not draw box outlines. Did not check if -D was not given, and had wrong test for -E. Added -Aa and -Al options to move only the annotations or label to the other side of the color bar. Now requires -Ac to keep writing the vertical labels as columns.
pstext.c
: We incorrectly removed blank lines but those mark new paragraphs when in -M mode.
psxyz.c
: Did not pick up y-size for column from input data file.
xyz2grd.c
: For -E under Windows we used fscanf with a pointer from GMT DLL which would fail for mysterious reasons.
meca/psmeca.c
: Fixed bugs that would give strange beach balls for some input.
meca/pscoupe.c
: Fixed bugs that would give strange beach balls for some input.
misc/gmtstitch.c
: Check to see if format was set was wrong.
mgd77/mgd77.c
: Index array error resulting in wrong IGRF start and stop years was fixed. Now applies recalculation of fields requested by E77 flags as part of reading netcdf mgd77+ files. However, if the original anomaly was NaN then we leave it as is.
mgd77/mgd77sniffer.c
: Bitwise assignment error was deactivating other fields when depth field was missing from a cruise. Another bitwise error was overwriting E77 nav flags when navigation was found on land Updated sample grid function to handle longitudes for img files.
mgd77/mgd77track.c
: Missing newline after last source line.
spotter/backtracker.c
: The -W option always assumed the reverse rotation, i.e., from hotspot to seamount. Now obeys the -D option.
x2sys/x2sys_get.c
: Now handles -R with longitude periodicity correctly.
xgrid/xGridEdit.c
: Needed GMT_io_init to get all pieces needed to read grids.

Here is a list of the recent enhancement to various programs:

  1. gmt_init.c has added support for enhanced -E option for 3-D perspective views which allows the specification of a fixed point (needed for creating animations). Updated all programs to use the new option and added updated man page and synopsis to all programs. Now, -R may take the name of an existing grid file. Then, the grid domain is used to set -R as well as the grid increment (and registration) for those programs that have such options.
  2. gmt_map.c also has added support for enhanced -E option for 3-D perspective views.
  3. gmtmath.c has added new operators NOT and INRANGE.
  4. grd2cpt.c can now accept multiple grid files at once.
  5. grdfilter.c has new options -Np to honor any NaNs found so output can be NaN, -Nr to replace output node with NaN if input node is NaN, and -Ni to ignore NaNs [Default].
  6. mapproject.c has extended the -A option by making the fixed point optional; if not given we compute azimuths between successive data points.
  7. minmax.c has optional /col that may be added to -Tdz to select another column [third]. Added -S to leave space for error bars. Useful with -I and subsequent psxy -E.
  8. pscoast.c has a new modifier + to the -D option, which determines the next lower resolution should the selected one not be available. This enhancement also affects both gmtselect.c and grdlandmask.c.
  9. psimage.c has new -Gt option, with assignment of color to be made transparent.
  10. pslegend.c lets N (number of columns) also affect the printing of labels. Can now use rectangle among the symbols.
  11. psmask.c now has modifiers +nn_points and +q to the -D option to limit the minimum number of points a polygon must have and, to suppress PostScript output, respectively.
  12. pstext.c has enhanced -Z+ option expects z-level values in 3rd column.
  13. pslib.c internals now measure paper size in double precision points instead of truncating to nearest integer. To remain backwards compatible for pslib.c users we now initialize all plotting in GMT with ps_plotinit_hires instead of ps_plotinit. Allow transparency when plotting 8-bit images (as well as 24-bit). We now use a new implementation of ps_textdim to ensure proper alignment of texts and the optional surrounding boxes. Finally, wehave simplified the PostScript code for symbols and removed some limitations on plotting by officially moving to PostScript language level 2.
  14. ps2raster.c now will scan for the optional comment %%HiResBoundingBox which takes precedence over the values in %%BoundingBox. New format -TG turns on transparency for PNG output, and -Tb selects Microsoft BMP output. Formats b, g, j, and t accept modifier “-” to produce grayscale images. Added -Cgs-command to pass one or more custom switches directly to ghostscript, and -Q[g|t]bits to set the level of anti-aliasing for graphics and text, respectively. We also added a new option -F to force a specified output file name. Finally, added -W to help create world files and geotiff output. To simplify boundary annotations for such plots (which must be inside the map region) we added the new choice inside for the BASEMAP_TYPE default parameter.
  15. psxy.c has new option -Iintens to modulate fill color via a fixed illumination value.
  16. psxyz.c has the same new option -Iintens.
  17. xyz2grd.c has new format (A) for -Z which allows more than one floating point value per input record. Cannot be used if the z-values are in dateTclock or ddd:mm:ss format.
  18. mgd77/mgd77track.c has enhanced -A option to place cruise ID equidistantly (distance or time) along the track.
  19. spotter/backtracker.c has new option -e to specify a single fixed total reconstruction rotation that will be applied to all input points.
  20. x2sys/x2sys.c internals now has automatic swabbing of index files, if required. Can now handle netCDF 1-D COARDS files.
  21. x2sys/x2sys_init.c Now, the distance and speed unit settings (-C, -N) are set here and kept with the TAG for use in other x2sys programs.
  22. x2sys/x2sys_datalist.c now has [experimental] support for using a correction table and can compute auxiliary data such as distance and azimuth.

Finally, we have added three new examples to demonstrate plotting of *.img grids, mixing UTM grids and geographic projections, and using greenspline.c for gridding on a spherical surface.

1.1.20 Overview of GMT 4.3.1 [May-15, 2008]

This quick update only 2 weeks after the release of version 4.3.0 was prompted by the discovery of three serious bugs; two of which were quite old but had caused no harm until tested under Fedora 9. The third critical bug prevented the wholesale reading and writing of GRD98 format grids. In addition a few minor bugs were discovered; this is the list of all corrections:

gmt_io.c
: GMT_nc_input would not read all data columns when no variable names were given.
gmt_mgg_header2.c
: Passed two of the arguments to fread in the wrong order. Prior to version 4.2.0 the return code (which indicated an error) was not checked yet the read did return the correct data. With better error checking this latent bug now caused a refusal to read any GRD98 grid.
gmt_support.c
: Allow a little more slop in determining whether primary tick is at same location as secondary tick.
gmtselect.c
: Did not like a variable number of input columns. Now OK if both input and output is ASCII and there are at least 2 (3 with -Z) input columns present.
grd2xyz.c
: For option -E, the test for xinc == yinc was susceptible to round-off.
project.c
: Now explicitly initializes the pointers in the data structure to NULL since realloc does not initialize new memory (yet almost all implementations of realloc appear to have done so anyway, masking the memory bug).
psxy.c
: The -D option erroneously gave an error despite being used properly.
psxyz.c
: Similar problem as described for project.
xyz2grd.c
: The -S option incorrectly insisted that -G must be used.

In addition, many of the supplements did not work properly under Windows due to internal problems with the DLL. Finally, one enhancement snuck in before the decision to issue this update was made:

  1. gmt_init.c was enhanced so that the media size Custom_WxH can use W and H in inch, cm, or m by appending i, c, or m to each dimension [Default remains points].

1.1.21 Overview of GMT 4.3.0 [May-1, 2008]

Changes are once again a mix of structural improvements, bug fixes, and a few enhancements. The coastline files (now GSHHS 1.10) have seen minor modifications, the mex supplement now offers support for Octave, all source code is now fully 64-bit compliant, we have added an isolation mode option (if GMT_TMPDIR is defined, write temporary and hidden files to that directory), and the configure/make setup has been further improved (such as honoring CFLAGS and LDFLAGS set by user). Colors may now be specified as hexadecimal codes (e.g., #ff0000 for red), and projections can be specified by name (similar to Proj4). Finally, binary table data can now be COARDS-compliant netCDF files. As for documentation, we have now switched from C shell to Bourne shell (although the csh examples are still distributed).

The following lists specific enhancements or new program options:

  1. gmt_grdio.c is modified so the grid i/o supports the GMT_[DATA|IMG|GRID]DIR environment settings.
  2. gmt_init.c was enhanced so -U can now interpret a justification (e.g., just/dx/dy on the command line or by setting UNIX_TIME_POS) and we introduced a new default parameter UNIX_TIME_FORMAT which controls the formatting of the timestamp.
  3. gmt_io.c now implements -b[i|o]c[var1/...] option to indicate input is netCDF.
  4. gmtmath recognizes new constants TMIN, TMAX, TINC, and N.
  5. grdimage uses -N to not clip image at map boundary.
  6. grdview now uses -Wf to change the facade pen from its default value.
  7. grdmath recognizes new constants XMIN, XMAX, XINC, NX, and similarly YMIN, YMAX, YINC, NY.
  8. mapproject -G+ will compute distances between coordinates in first 4 columns.
  9. ps2raster has new option -D to specify alternative output directory and -V to report progress.
  10. psrose has new option -F to disable the plotting of the scale bar.
  11. psxyz has new option -D to match option set in psxy.
  12. mgd77/mgd77list added -Ga|brec to limit output to a certain record range.
  13. spotter/hotspotter added -S to normalize output to percent of CVA maximum.
  14. spotter/grdspotter is a new program, like hotspotter, but using gridded data as input.

A long list of bugs has been squashed since the last release, the most important are listed below:

gmt_grdio.c
: Fixed 3 bugs in GMT_decode_grd_h_info that caused problems parsing -D option. Explicitly exclude = sign from becoming separator.
gmt_io.c
: Now skips blank lines that has leading whitespace.
gmt_init.c
: c for seconds was not recognized as TIME_UNIT (expected s, which is kept for backwards compatibility). -B processing of labels used an internal string that was too short, which could lead to label truncation. Fixed “Holiday-bug” in GMT_parse_J_option introduced 2007-12-21. GMT_str_tolower could run out of bounds. Probably only affected 32-bit compilers. GMT_is_a_blank_line is now used wherever ascii input is processed.
gmt_map.c
: Fixed bug in radial clipping. The radial clipping would sometimes add arcs using the arc that exceeds 180 degrees. Added new rectangular clip function using Sutherland/Hodgman algorithm in order to fix incorrect results in grdlandmask. Minor bug in 4th term in conformal to geodetic lat. Round-off could mess mapping of west/east to xmin/xmax. Now has a safety valve for checking that this does not occur.
gmt_plot.c
: 3-D basemap axis did not use LABEL_OFFSET.
gmt_proj.c
: Fixed bug in azimuthal equal area projection that had the horizon shifted from where it ought to be. Avoid error in GMT_lamb_sph when lat is 90 degrees. Clip path for general perspective projection was not closed. For -JS: Would set slice to NaN if central meridian was not Greenwich. For -JR: Longitudes beyond 180 were set to 180.
gmt_stat.c
: Bug in GMT_median would sometime give subtle mistakes, most noticeable when only a few values were passed to the function. Traced to the use of size_t variables in expressions that could yield a negative value. Fixed minor issues in GMT_PvQv function.
gmt_support.c
: Made GMT_polygon_is_open tolerant of round-off and if polygon is not open set last to exactly equal first point. GMT_get_annot_label did not properly honor the ddd.xx setting. Now implements annotation for Gnomonic maps. In GMT_contour, would occasionally not check internal crossings for some interior contours.
pslib.c
: Bug in ps_shorten_path lead to SEGV when path resulted in a single point. Fixed error in applying pstext’s -Djdx/dy shift in paragraph mode (-M). Redefine rect symbol to be less prone to round-off. Used internal point_code before it was initialized. This caused PAGE_COLOR not to work (wrote <NUL > rather than C).
gmtmath.c
: Implemented Welford (1962) algorithm in KURT, SKEW and STD operators for more precise one-pass computation of mean and sum of squares.
gmtselect.c
: The -:o option failed to reverse output order.
grd2xyz.c
: Do not abort when -R exceeds grid; simply output common region. For gridline oriented grids: -E returned xll and yll one cell too large. Now writing [xy]llcenter properly.
grdblend.c
: Did not pick up node registration before calculating output grid size.
grdedit.c
: Adjust z_min and z_max when changing add_offset or scale_factor.
grdgradient.c
: With -D and -S the slopes were not set to NaN if data were NaN.
grdinfo.c
: Implemented Welford (1962) algorithm for more precise one-pass computation of mean and sum of squares. zmin==zmax no longer forces -M option.
grdmask.c
: Tiny bug for determining which hemisphere (N/S) unlikely to have had any effect. Needed to allow for some slop when comparing shrink to 0.0 since sometimes the result of acos is 1e-14 or thereabouts.
grdmath.c
: Implemented Welford (1962) algorithm in KURT, SKEW and STD operators for more precise one-pass computation of mean and sum of squares.
grdview.c
: The facade (-N) outline was drawn with contour pen. The -Qc option failed to set the “build image” flag and produced garbage surface tiles. Also, -T[s] produced polygons that were not checked for wrapping at a periodic map boundary. Clarified that -T cannot take -JZ|z.
minmax.c
: Could get confused when longitudes crossed dateline or Greenwich, and OUTPUT_DEGREE_FORMAT could interfere with result.
pscontour.c
: -T option was susceptible to infinite loop if bad record was found.
pshistogram.c
: The -R option was processed separately and did not understand time coordinates. Bug in -F option failed to center bins.
psmask.c
: The -D option used the wrong output file name.
psscale.c
: Inverted vertical scale, when using filled rectangles: Colors remained in the original order. Inverted vertical or horizontal scale, when using rectangles with gradients: Size of rectangles followed original order, not inverted. In reverse mode, -Eb was plotting foreground triangle, -Ef background triangle.
pstext.c
: Parsing of -C complained about % sign.
psxy.c
: If first symbol in list with size was not circle, it got rejected. When sizes of -Sr or -Sj were read from list, they were always assumed to be in inches. Drawing arrows with -SvS and time-coordinates did not work as 2nd time coordinate did not get processed properly.
psxyz.c
: If first symbol in list with size was not circle, it got rejected. When sizes of -Sr or -Sj were read from list, they were always assumed to be in inches.
sample1d.c
: Calendar time knots did not get properly interpreted with -N.
mgg/mgd77togmt.c
: Did not initialize the MGG_SHAREDIR path.
mgd77/mgd77.c
: Wrong header order written if mgd77convert ... -Tt was used.
mgd77/mgd77info.c
: Could get confused when longitudes crossed dateline or Greenwich, and OUTPUT_DEGREE_FORMAT could interfere with result.
mgd77/mgd77sniffer.c
: Numerous fixes and enhancements; see ChangeLog.
x2sys/x2sys.c
: Error in determining which columns had been requested.
x2sys/x2sys_binlist.c
: Could create bad bins because of incorrect reallocation of memory.
x2sys/x2sys_get.c
: The -L option did not honor any -F or -N settings.
x2sys/x2sys_put.c
: Wrong test when replacing older track info lead to data base loss.
x2sys/x2sys_cross.c
: Used wrong data column order and computed speed when there is no time.

1.1.22 Overview of GMT 4.2.1 [October-10, 2007]

Changes in GMT 4.2.1 once again address many structural issues as well as numerous bug fixes. System-wide changes include a revamping of the entire configure/make setup for both regular installations and CVS users, an improvement to how the BCR 2-D interpolations for images and grids are done by adding B-spline and nearest neighbor as optional interpolants, introduction of a new PostScript Level 2 pattern machinery in pslib.c, an updated GSHHS coastline version (which also includes Australia internal state boundaries, fixes to the Yemeni and Lebanese borders, and more river lines), and general improvements and corrections to the documentation, such as placing all man pages in section 1 (except pslib which goes in section 3). Starting with GMT 4.2.1 we will also begin naming the coastline-related archives by the GSHHS prefix and use the actual GSHHS version number (now 1.9).

Individual programs have also seen some new options or enhancements:

  1. gmtselect has enhancement -Lp to limit points beyond a line’s endpoints.
  2. grdfilter has new option -D5 to allow direct filtering of Mercator grids (img).
  3. grdmask has an enhanced -A option; append m or p to design a mask polygon by first following a meridian, then a parallel, or vice versa.
  4. gmtmath has several new operators, such as PLMg (geophysical normalization and suitable for high degree and order), FACT, SKEW, KURT, PQUANT, EULER, PSI, PV, QV, COT, COTD, ACOT, SEC, SECD, ASEC, CSC, CSCD, and ACSC.
  5. grdmath has the same new operator as gmtmath, plus YLMg and CBAZ.
  6. grdproject now considers -R an optional setting.
  7. mapproject has an enhanced -G- option where increments rather than cumulative distances are returned. Also, for UTM projections with -C, the -R option is now optional provided the UTM zone is properly specified.
  8. ps2raster has now a new -P option for forcing portrait orientation.
  9. pshistogram has now a new -C option for using a cpt file to paint the bars.
  10. pstext will now accept the @_, @: and @; escape sequences for underline, font size change, and font color change regardless of mode.
  11. psrose has a new option -L to control the labels.
  12. psxy has an enhanced -A option; append m or p to draw a line by first following a meridian, then a parallel, or vice versa. Can now plot a notched box-and-whisker symbol, and we added +ndx/dy to nudge placement of quoted line labels. Encanced-E.../[-|+]pen where + means apply cpt color (-C) to symbol and bar - means apply to bar only (no fill). Likewise, -W[+|-]pen controls if -C sets outline (-) or both fill and outline (+).
  13. psxyz hass enhanced -W[+|-]pen that controls if -C sets outline (-) or both fill and outline (+).
  14. sample1d now accepts -Fn for no interpolation (returns nearest value).
  15. Two additional color maps have been added: panoply mimics the default colormap in the netCDF viewing program Panoply; cyclic provides a full spectrum of 360 degrees in hue.
  16. originator (spotter supplement) has new option -Q to specify constant r/t for (x,y,z) data only.

Below is a list of previous problems that we have identified and corrected in the current release:

gmt_agc_io.c
: Failed when nx and/or ny was multiple of 40 + 1.
gmt_custom_io.c
: Surfer grid start at lower, not upper left.
gmt_io.c
: Combinations of OUTPUT_DEGREE_FORMAT=ddd:mm:ss.xxxF and -: would add W|E to lat and S|N to lon for formatted output. Blank lines were not recognized under cygwin/SFU. When no delimiters are used for input date format (e.g., yyyymmdd) then yyyy must be 4 characters and we must use %4d. Leading zeros are required if year < 1000.
gmt_init.c
: For quoted lines, the :Lh modifier did not reset label if a segment header had no label specified. Also, label would include the leading ïn a multi-word label. Now, UTM zones may be A,B,Y,Z or 1-60 with modifiers C-X (except I and O). Parsing of psxy’s -Sf incorrectly scaled a count to inches. Suffices h|+|- in -J were mistaken for units.
gmt_map.c
: KM_PER_DEG was not reset in GMT_set_spherical. This may have caused errors in scaling when ELLIPSOID was not set to Sphere. GMT_distances function did not work with correct pointer. Great circle intersection did not select correct vector sign so points 180 degrees from a line could pass as close in gmtselect. Removed restriction that east and west limits can not be both negative. This fixed the remaining Hexagone problem.
gmt_nc.c
: The attribute actual_range was erroneously stored in grid units in stead of actual units. The difference is only relevant when storing scaled integers. A backward compatibility for grids written with previous versions is built in. Also the GMT version number is added to the global attributes and text attributes are truncated to their proper length.
gmt_plot.c
: Failed to activate ddd.mm.xxx format for maps. Oblique tickmarks sometimes were missing for latitudes.
gmt_proj.c
: -JS inverse did not apply quadrant check.
gmt_support.c
: ddd:mm.xx (2 or more x) failed to format properly. Needed to reverse z_low and z_high values as well as rgb values when reversing color maps.
gmt_vector.c
: Function GMT_fix_up_path: Intermediate longitudes now wrapped based on segment, not on map extent.“greenwich” argument became superfluous and was removed. Argument “step” was found to be in degrees, not in inches as some calling programs assumed.
filter1d.c
: For -Fffile, set filter width to DBL_MAX until filter is read.
grd2cpt.c
: -E flag did not work unless -T was used.
grd2xyz.c
: Did not switch to binary i/o mode when -Z specified binary output (Windows bug only).
grdblend.c
: Now works with Windows DLL by using the GMT i/o functions.
grdcontour.c
: -Avalue would turn off the expected default transparency.
grdcut.c
: Domain check failed for geographic 360-degree grid with -fg.
grdedit.c
: Disallow -T for Surfer grids since they don’t support both kinds of node registration. -N did not take into account if -fg was given.
grdfilter.c
: -Inx+/ny+ was not propagated to the output grid setup.
grdimage.c
: Removed -T option; no more polygon drawing (use grdview). -Smax_radius changed to -S[-]b|c|l|n[/threshold] to take advantage of new and improved projection function GMT_grd_project.
grdlandmask.c
: Failed to set rightmost, empty bin if 360-periodicity in effect; also needed to set repeating right column to left column.
grdmask.c
: For -S..k, did not properly account for latitude effect on dx. SEGV error when GMT_fix_up_path returned f ewer points than originally allocated and the n_alloc variable was not reduced accordingly.
grdmath.c
: When M=0, set Imaginary component of YLM to 0 (was same as real).
grdproject.c
: -Smax_radius changed to -S[-]b|c|l|n[/threshold] to take advantage of new and improved projection function GMT_grd_project.
grdreformat.c
: Could not determine format of output file automatically since file does not yet exist (must append file-type code).
grdsample.c
: -Qvalue changed to -Q[-]b|c|l|n[/threshold] to take advantage of new BCR code.
makecpt.c
: Colormaps with non-equidistant intervals can now be reversed properly.
project.c
: Wrong azimuth for -N and -G generating lines.
pslegend.c
: Usage message did not explain the -L option. Added @ECHO OFF to avoid commands being echoed into the PostScript output under Windows. The color change (C) macro messed up -Xa and -Ya absolute settings.
psbasemap.c
: Check that -L is used with geographic coordinates only.
psimage.c
: The -I option had become always active.
pslib.c
: Now no text is written if font size equals 0.
psscale.c
: Annotations did not work properly if -Q and -B1p were used together.
psxyz.c
: Fixed bug in painting outline of text symbols.
xyz2grd.c
: -E did not process pixel grids properly.

A few bug-fixes applies to the supplements as well:

gshhstograss.c
: Removed use of getopt so even the lamest systems can compile it (incuding Windows).
img2mercgrd.c
: Fixed AND vs OR logic in lat bounds check when -D was set. Used wrong origin when -C was used. Now relative to lon = lat = 0.
DLL
: The meca supplements now work with Windows DLL by using the GMT i/o functions.
utilmeca.c
: Make sure that checks on floating point numbers work well, even in case of small round off errors. Thanks to Peter Lombard.
mgd77manage.c
: -Qvalue changed to -Q[-]b|c|l|n[/threshold] to take advantage of new BCR code. Added E77 status attribute to MGD77+ files.
mgd77sniffer.c
: Same. Adjusted anomaly recalculation code so that cruises with m=1 and b=0 are reported to be same as expected. Repaired 8 memory leaks. Fixed two regression bugs, one in grid comparisons in which not all data were copied into the regression arrays and the other in faa recomputation regression where eot was being applied twice. Added code to check if regression is outside a specified percent (-P). Updated e77 messages and moved along-track grid offset errors to e77 header rather than having its own error class.
gmtdigitize.c
: Must use separate x and y-scales if -Jx is specified. Wrong coordinates used to calculate rms misfit.
originator.c Now lists option -L in both usage and man page.
x2sys.c x2sys_set_system did not initialize structures properly.
x2sys_cross.c Skip duplicate files with a warning. Distance calculations were bogus (see gmt_map.c), and wrong number of arguments passed to readfile function pointer.

Finally, as far as CVS users are concerned, the old "gurumake" system has gone. To compile from CVS, users need to use a GNU compatible make program. A combination of GNUmakefile and makefile files make sure that those components not in the tarballs are created from scratch. Type make in the GMT directory to get a list of targets.

1.1.23 Overview of GMT 4.2.0 [April-1, 2007]

Changes in GMT 4.2.0 address many structural issues as well as many bug fixes. We have consolidated user initialization files in the  /.gmt directory, continued to replace tiling with bitmaps, and have performed a myriad of under-the-hood changes. One imporant and more visible new feature is the fact that grdimage and pscoast now can use the general perspective projection with arbitrary elevation (-JG has been enhanced to handle the extra arguments required – see the new example 26 for details). Also, the coastline files have been updated to use GSHHS version 1.5 which fixes minor inconsistencies in the coastline database. We have also corrected issues that made the Windows DLL explode in 4.1.4. Finally, a few enhancements have been made to these programs:

  1. NGDC’s GRD98 format has been updated to handle both gridline and pixel node registrations.
  2. We have relaxed the restriction on latitude for -JA, -JS for polar aspects; now more than one hemisphere may be displayed. Better warning/error messages.
  3. gmtconvert has an enhanced -E option; append f or l to only get first or last record per segment.
  4. gmtmath -T can now have + appended to indicate number of points instead of increment.
  5. grdcontour has a new option -F to orient dumped contours. Can now append :radius to the -G option in order to specify a minimum spacing (measured in the x/y plane) between contour labels.
  6. grdinfo has an enhanced -I option. With no arguments we return the grid’s -Idx/dy string whereas -I- will return the grid’s -Rw/e/s/n string.
  7. grdmath has new option -M for using map units in gradients and new D2DXY operator. Also added SBAZ for back-azimuths and now allow ELLIPSOID to control if great circles or geodesics should be used (Sphere selects great circles)
  8. psrose has new -D option to center the sector bins (like pshistogram -C).
  9. psxy understand -W- and -W+ in multisegment headers which will turn off outline or reset to default, respectively. Similarly, -G- and -G+ will turn off fill or reset to default (with -M). Also added new option -SB for horizontal bar (-Sb is vertical).
  10. psxyz also has -SB for horizontal bar (-Sb is vertical).
  11. sample1d now allows absolute time in -S option.
  12. imgsrc/img2mercgrd can take -C to let the Mercator x/y use the global origin of img file.
  13. Because of its popularity, ease of use, and importance to many user how otherwise would not know about its existence, ps2raster is moved from the supplementary misc directory to the main set of GMT programs.

Below is a list of previous problems that we have identified and corrected in the current release:

gmt_init.c
: Fixed unit problem with-Jx1:xxxxx. Erroneously added degree symbol to both coordinates in case of -JX..d/.. (single d). Now properly adds only degree symbol on specified axis. Did not change time-system when only TIME_UNIT was specified. Failed to properly parse a single PAR=ARG (one word) argument given to gmtset.
gmt_io.c
: Did not terminate a calendar string after copying it.
gmt_plot.c
: The logic to check for seconds annotations failed if inc <1 arc second. Did not set the contour annotation font before writing labels.
gmt_support.c
: Parsing old-style pens did not set offset to 0 when no texture was given. Fixed IFACT size in the old Brenner FORTRAN FFT – bug undetected since GMT 1! Contouring of grids with NaNs need to check both vertical AND horizontal interior gridcell boundaries for possible crossings. The label machinery for ddd:mm:ss.xx used the wrong parameter to check for fractional seconds annotation (the .xxx part).
gmt_time_systems.h
: J2000 epoch was 1.0 Jan 2000, instead of 1.5 Jan 2000.
filter1d
: The -T option can now parse datestrings for the min/max fields.
gmtmath
: Fixed memory allocation bug for files with more than BUFSIZ records. Now works correctly with multisegment headers. Multisegment headers now written to the output file and not always to stdout. Option -C now works (used to deselect all columns). Fixed LSQFIT (used wrong columns when some were skipped).
grdcontour
: Interior contours were not smoothed unless NaNs were involved. Also the labeling of closed highs/lows were insensitive to pixel versus gridline grids and could get the wrong result. -C need to check for“.cpt” at END of file name. Fixed contour label angle specifications were always ignored.
grdfft
: -D and -I options could have junk in the parameter arrays if given more than once.
grdimage
: -JX with a negative scale/length and -Edpi failed to flip the image.
grdinfo
: Would not take both -L1 and -L2.
grdview
: Plot no mesh when -T is used.
mapproject
: -L option did not allocate enough output memory for extra columns.
pscoast
: Did not list -Z in synopsis/usage.
pscontour
: For 3-D views, -E projected contours (-W) twice and mesh lines (-L) not at all. Fixed contour label angles were always ignored.
pslib.c
: Only issue setdash PostScript commands if texture has changed. ps_clipoff needed to reset memory of last pen width/color/pattern. ps_color_tiles used wrong node registration.
psrose
: Failed to skip pie-slice filling when no fill was selected. -C lead to SEGV as it tried to read from a non-existent file.
psxy
: When used with -S but no -W or -G given, only set default -W if -M not used. Erroneously turned symbol outlines on if just -C was given. Incorrectly suggested that -Svs needs 5 instead of 4 columns. -Sf option was parsed to require 3 instead of 2 coordinates.
psxyz
: -Svs could fail to plot by confusing y-coordinate and y-size. Also need 6 rather than 5 input columns. Bar width has only half of what was requested
splitxyz
: Reported headings in radians instead of degrees.
trend2d
: Failed to pass the new variable with column choices. Also did not accept z as an output choice.

A few bug-fixes applies to the supplements as well:

gshhs.c
: Had && instead of & in bit-arithmetic that reported level.
mgd77manage
: Failed to enforce that a new column abbreviation must be in lower case.
x2sys_cross
: Failed to check for crossovers falling exacly on data nodes.

1.1.24 Overview of GMT 4.1.4 [Nov-1, 2006]

Changes in GMT 4.1.4 are again relatively minor and predominantly bug fixes. One imporant new feature is that GMT can now automatically recognize the format of the grid file given to a program. The use of the “=id” mechanism is now only needed when writing an output file in a grid format other than the netCDF default or when reading using custom scaline and translation is required. We have also added a new user directory pointed to by GMT_USERDIR (default directory is ~/.gmt) where items such as .gmtdefaults4 will be looked for. Additionally, a few enhancements have been made to overcome limitations in the previous versions:

  1. grd2cpt has a new option -T for the creation of tables symmetric about zero.
  2. grdblend will accept negative weights which are taken to mean that the sense of tapering should be reversed.
  3. grdedit has a new option -E to transpose the entire grid.
  4. grdmath has a new option -N to turn off strict domain match checking when multiple grid files are involved.
  5. grdreformat now supports the -f option.
  6. nearneighbor will now optionally accept a min_sectors argument appended to the -N option.
  7. pshistogram’s option -I can now accept a modifier O to output all bin data even if y = 0.
  8. psscale will now invert the color scale if a negative length is provided, and -I will now work with colormaps with non-constant interval and with gaps (-L).
  9. psxy and psxyz have a new option -Sj|J that plots a rotatable rectangle but otherwise behaves similarly to -Se|E.
  10. ps2raster has many improvements; added EPS output; high-quality PDF output. Also removed -dDOINTERPOLATE option which caused inversion of colour map and had no benefits.

Below is a list of previous problems (a few accidently introduced in GMT 4.1.3) that we have identified and corrected in the current release:

gmt_agc.c
: AGC grids use 0 to represent NaNs – this was not implemented yet.
gmt_calclock.c
: Proper rounding of time when converting to dates.
gmt_support.c
: Fixed bug in -I when modifier = was used.
gmt_init.c
: Fixed bug not recognizing PAGE_ORIENTATION as well as a bug that prevented proper writing of PAGE_ORIENTATION in defaults. Added a check so gmtset will not crash if VALUE is not given. Finally, let GMT_HOMEDIR default to C: under Windows if HOME is not set.
gmt_io.c
: GMT_scanf_argtime now returns RELTIME (not ABSTIME) when relative time is found.
gmt_grdio.c
: Set [xy]_units also in GMT_update_grd_info. Fixed time-scaling bug on input (was always seconds). Make units check case-insensitive. Check for toggled lat/lon coordinates.
gmt_nc.c
: Made sure no garbage remains under Cygwin when using strncpy. Check if x- and y-coordinates have constant step sizes; issue warning if not.
gmt_plot.c
: Bugs related to annotations with -JPa and its z modifier fixed. Log gridlines did not work for 3-D view. 3-D axis label would sometimes get misplaced due to round-off. 3-D map scale did not project correctly. Duplicate title could appear if -JX was used and one axis was geographic (d). Needed to add secondary font to list to be encoded.
pslib.c
: Fixed memory management in LZW compression (memory leak). Improved EPS conformance.
filter1d
: Robust option used extreme rather than median to determine the outliers.
gmtconvert
: Did not have -L listed in synopsis.
grdblend
: Now skip grids that are entirely outside the region of interest.
grdcontour
: Crashed if -M and -D were used with no file name specified. The -W[+][c|a] option was susceptible to misinterpreting things like -Wcyan as contour pen with color yan.
grdcut
: Require geographical instead of global in order to shift by 360 degrees.
grdfilter
: Should not wrap over pole unless grid extends all the way to the pole.
grdinfo
: When -C was used there was no linefeed at the end.
grdsample
: -T did not ignore -R (as per manual), resulting in changed cell size. -F did not use gridline node registration as default, rather that of the input grid. When using pixel node registration, number of cells would be one too large. -L worked only in very limited case: going from x=[-180;180] to x=[0;360]. Now supports any periodicity in X and Y (as per manual). -F again forces pixel node registration. Default is same as input. More consistency with manual.
grdtrack
: The -Z option failed to be set for some input configurations.
grdvector
: Added -f option.
grdview
: The -W[c|m] option was susceptible to misinterpreting things like -Wcyan as contour pen with color yan.
grdvolume
: Three bugs squashed: gridcell oriented grids now get proper area and volume, including edges; only one cell per NaN is excluded; when -C and -L are combined, the volume is properly corrected for the baseline height.
pscoast
: -N and -I reset pens to default settings after initially changing them. Did not change output mode to binary (Windows only) if -M and -b were set. Both -Gc and -Sc needed to check that no letter is following the c modifier.
pscontour
: The -D dump option wrote projected instead of original coordinates.
psimage
: -Gcolorname (e.g., -Gblack) will now be interpreted as foreground color, e.g. -Gfblack.
pslegend
: Did not replace octagons with polygon form when pattern was requested. Did not consider if absolute coordinates were given in -X and -Y. Passed the wrong character code when M was chosen with a plain scale modifier.
psscale
: A vertical bar with a label placed along it was mis-justified.
pstext
: Default for -G is now BASEMAP_FRAME_RGB as for other map annotations. The box option -W[fill][o|O|c|C][pen]] is now -W[fill,][o|O|c|C][pen]], i.e., we now use a comma to separate the fill and pen (done since fill may be a name containing o|c); parsing is backwards compatible.
psxyz
: -C for symbols did not pick up color fill.
trend2d
: Processing of -F happened after checking.
xyz2grd
: Had -Az as default rather than no -A. Fixed bad header parsing when -E was selected.

A few bug-fixes applies to the supplements as well:

grdraster
: Only do 360-degree wrapping if working on a geographic grid.
mgd77list
: Did not process time when -Am2|4 was set and time was not requested as output. Also, did not process time when -Am2|4 was set and time was not requested as output.
x2sys.c
: Did not look in current dir for *.def files.

1.1.25 Overview of GMT 4.1.3 [June-1, 2006]

Changes in GMT 4.1.3 are relatively minor and predominantly bug fixes. However, a few enhancements have been made to overcome limitations in the previous versions:

  1. Added the Hughes 1980 ellipsoid for projection support for DMSP SSM/I grid products.
  2. grdfft has an extended -F option to allow for either Gaussian- or cosine-tapered filtering.
  3. psscale now has a -Q option so that logarithmic color scales and annotations can be handled properly.
  4. makecpt and grd2cpt have a new -M option to allow the background, foreground, and NaN-colors to be assigned using the GMT defaults instead of the settings in the master CPT file.
  5. mgd77list in the mgd77 supplement has new option -Q to specify limits on speed and azimuths for output records.

Below is a list of previous problems (some accidentily introduced in GMT 4.1.2) that we have identified and corrected in the current release:

gmt_grdio.c
: Bug in GMT_grd_shift for gridline-registered grids; this function is used in grdedit to rotate grids of 360-degree longitudinal extent. Also added better testing for subsets of global (0-360) grids.
gmt_init.c
: GMT_PS_init was called after –PAR=val had been decoded, resetting the PostScript-related parameters to their default settings.
gmt_support.c
: GMT_set_xy_domain padded region for pixel instead of gridline node registration, which could cause SEGV in xyz2grd if (x,y) was less than half the grid-spacing outside region.
blockmean
: The -C option got reversed in 4.1.2 - now fixed.
blockmedian
: The -C option got reversed in 4.1.2 - now fixed.
grdcontour
: The -C option with a non-cpt file failed to read due to lack of proper if-test.
grdedit
: The -S option was backwards and tested w-e=360; should be e-w=360.
grdimage
: Fixed bug introduced by GMT_get_inc in 4.1.2. Internal projected grid never took node_offset from input grid.
grdmask
: Polygons with zig-zag shape would sometimes cause a node exactly on a polygon vertex to be considered inside. Radius was reset to 0 after -Sradius was assigned.
grdvector
: The -A option was not properly initiated.
psbasemap
: The -L option did not properly parse the optional :label:<just> part.
pslegend
: If the M (for map scale) was used, the -R and -J options would be reset. Also prevented the undoing of -X and -Y at the end of the program.

1.1.26 Overview of GMT 4.1.2 [May-15, 2006]

On the surface, changes in GMT 4.1.2 are relatively minor. Most of the work has involved realignment of code and parsing of arguments to simplify the upcoming port to GMT 5; a brief listing of more visible changes include

  1. Coastline files have been updated to use GSHHS version 1.4 which fixes minor inconsistencies in the coastline database.
  2. All coastline files are now stored in a new subdirectory coast under the share directory, and the tar archives for coastlines now have their own version numbers as they do not change as frequently as the source code. Current coastline version number is 4.1.
  3. The archives have been reorganized so that GMT_share.* contains all files needed at runtime whereas the standard coastline files are in the new GMT_coast.* archive. The GMT_progs.* archive has been renamed GMT_src.*.
  4. CPT files can now have z-values that are date-time strings.
  5. Optionally append z to the -Jp projection to annotate depths (i.e., “north-y”) rather than radius.
  6. Two new tools added to the misc supplement for digitizing lines (gmtdigitize) and to stitch digitized lines into continuous lines or polygons (gmtstitch).
  7. Extended -M option to take optional modifiers i or o for input or output.
  8. Added support for custom grd format AGC from Atlantic Geoscience Center, assigned the code af [21].

A few programs or options have received minor updates and new features, such as

blockmean
: Added -E for reporting standard deviation, min, and max values per block.
blockmedian
: Added -E for reporting L1 scale, min, and max values per block. Also added -T to specify a particular quartile [Default q = 0.5 = median].
blockmode
: Added -E for reporting LMS scale, min, and max values per block.
configure
: Added explicit options to bypass the installation of supplements that require Matlab (–disable-mex) and X11 (–disable-xgrid).
gmtconvert
: Added -D option to write segments to individual output files.
gmtdefaults
: Support for new default PS_VERBOSE which controls the writing of comments to PostScript files. COLOR_MODEL can now accept a prefix “+” which forces color interpolation in the selected system (RGB or HSV only). Default remains RGB. PS_COLOR has been extended to accept HSV as well (only applies to polygon, symbol, pen, and text colors, not images.). New parameter POLAR_CAP which controls the number of gridlines that converge on the poles for azimuthal and some other projections. Added new default HISTORY [TRUE] which controls whether or not we maintain a common command option history via .gmtcommands4 files.
gmtmath
: Added option -M to indicate the program can now handle multisegment files. Added CPOISS for cumulative Poisson distribution.
grdmath
: Added CPOISS for cumulative Poisson distribution.
minmax
: -D made obsolete by improved range checking for longitudes (but available for backwards compatibility).
psscale
: Enhanced -I option for asymmetrical intensity ranges from low to high.
psxy
: Added -SW for wedges defined by azimuths rather than directions. Polygons of large longitudinal extent now clip correctly.
splitxyz
: New option -Q to specify the output columns and their order.

Below is a list of previous problems that we have identified and corrected in the current release:

gmt_plot.c
: The 3-D perspective plotting of text was not scaled correctly.
gmt_support.c
: Parsing of -L option used in psbasemap and pscoast failed to get correct unit when ddd:mm:ss syntax was used for the position. Corner boundary conditions for grids (needed by grdtrack, grdsample, grdview, and grdgradient) had the wrong sign.
gmt2rgb
: Did not check name template properly, and did not initialize region.
gmtselect
: Option -F insisted on using spherical testing for Cartesian x,y data.
grd2xyz
: The -E option had the y-direction reversed.
grdfilter
: Needed the -f option to process -Rddd:mm syntax.
grdimage
: Would hang in stdin if -C not given when one grid is plotted.
grdmask
: Did not explicitly close polygons before using them. Test for polar caps applied to the opposite pole.
grdmath
: Command INSIDE for Cartesian data had bug (passed y where x was expected).
grdsample
: Failed when -I was specified.
grdview
: Bug in plotting north facade (-N). Also tried to free unallocated memory if -G was used.
project
: Cartesian projections gave incorrect results for p,w,r,s. Removed 0–360 restriction on azimuth. Option -G was susceptible to round-off and thus sometimes reissued the final point.
psxy
: -SV and -SE for -JX did not convert azimuths to directions. The -Sq option would get confused when distances between successive labels were smaller than the line’s point spacing.
mgd77/mgd77manage
: Did not properly close the file after ingesting E77 information.
pslib.c
: ps_load_raster did not use open mode rb and hence failed under Windows.
xyz2grd
: The -E option had the y-direction reversed.

A few bug-fixes applies to the supplements as well:

x2sys_get
: -N did not work properly (now fixed and tested).

1.1.27 Overview of GMT 4.1.1 [Mar-1, 2006]

Changes in GMT 4.1.1 are mostly minor; a brief listing include

  1. gmt_nc.c: Introduced handling of 4-D COARDS compliant grids (See Chapter 4 for details).
  2. mgd77/mngd77sniffer: New tool for along-track quality control checking of MGD77 files.
  3. spotter/grdrotater: New tool that rotates grids given a specified finite rotation.
  4. Jonathan Shewchuk’s triangulation routines are now stored with the rest of the source in the GMT_progs.tar|zip archives. (However, because his copyright is not GPL, installing it is still an option).

A few programs or options have received minor updates and new features, such as

grdedit
: Added option -T to toggle between gridline and pixel node registrations (header only).
grdgradient
: Implemented variation on Lambertian illumination.
grdmask
: Now takes -Sradius[c|m|k|K] as is done in nearneighbor.
gmtmath
: If file is STDIN we read data from stdin and put the contents on the stack. Also added -F to select which columns to use for output [all].
grdtrack
: Can now sample Sandwell/Smith IMG grids directly.
psmask
: -Now takes -Sradius[c|m|k|K] as is done in nearneighbor. Can now plot tiles regardless of projection and use patterns.
pstext
: -D[...]vpen can now be used with or without -M.
psxyz
: -SO|U imitate -So|u but without the 3-D color shading.
mgd77.c
: Added mechanism to search directories for files (mgd77 supplement).
mgd77list
: Activated -X option and associated machinery for applying data corrections (mgd77 supplement).

Inevitably, when new features are added, new bugs come along with them. Below is a list of problems that we have identified and corrected in the current release:.

configure.in
: Extracting VERSION from gmt_version.h, not gmt.h.
gmt_init.c
: BASEMAP_FRAME_RGB overrode any changes to grid pens etc. Now only does so if prefixed by ’+’.
gmt_calclock.c
: Did not allow -B0 for time-axis.
gmt_map.c
: -JX...d now plots with WESN or degrees:minutes as per PLOT_DEGREE_FORMAT. Map clip paths for -JElon/±90 were no good. Under certain circumstances, GMT_non_zero_winding might be passed a polygon that was not closed, resulting in an error. -JQ would give garbage if central lon was way outside -R.
gmt_plot.c
: -JX...d now plots with WESN or degrees:minutes as per PLOT_DEGREE_FORMAT.
gmt_grdio.c
: Changed logic to avoid false “scale==0” warning on Windows. GMT_open_grd (used in grdblend) reset scale to NaN. Initialize header information at start of GMT_read_grd_info.
gmt_support.c
: Initialize [xyz]_unit with more appropriate values. Got wrong conversion for dx in meters to degrees.
gmt_grd.h
: Improved definition of GMT_x_to_i macro should reduce bugs
pslib.c
: ps_polygon: if outline == -9 just fill and no clip. Fixed two bugs concerning the /MaskColor operator.
ISO-8859-9.ps
: Added /dotlessi per Onur Tan.
blockm*
: Now correctly deals with periodic longitude data.
grdcontour
: Fixed several issues at grid limits and inappropriate scaling of grid dimensions..
grdfilter
: Used -1 as index flag instead of INT_MIN.
grdimage
: Fixed several issues at grid limits and inappropriate scaling of grid dimensions.
grdmask
: Only let you change the value for outside nodes.
grdmath.man
: Did not list -f option. Operators LT, LE, EQ, GE, GT returned TRUE if NaNs were involved Now NaN is returned if any of the two operands is a NaN.
grdreformat
: Update grd.command before writing grid
grdvector
: Did not place vectors correctly for pixel-pregistered grids.
grdview
: Skipped nodes outside boundary but they might be needed to draw a tile.
pscoast
: With -JE and -Gr/g/b, the painting of the antipodal bin would incorrectly turn off clipping, messing up the rest of the plot. Now pass -9 to GMT_fill which means just fill and no end of clipping.
xyz2grd
: For geographic grids with 360° range and gridline node registration, the west and east bin did not get replicated properly. Now considers data inside the first and last tiles which might stick outside w/e/s/n.

A few bug-fixes applies to the supplements as well:

x2sys_cross
: Several problems fixed.

1.1.28 Overview of GMT 4.1 [Jan-7, 2006]

Most changes in GMT 4.1 are improvements “under the hood”. The most significant of these are

  1. Addition of ability to both read and write netCDF files that are COARDS compliant. This means that GMT, for the first time, is able to read netCDF files created by applications other than itself, and that other applications capable of reading COARDS-compliant netCDF grids can directly import data from GMT. We have added the new parameter GRID_FORMAT to the GMT defaults with “nf” as default. Users who, against our recommendation, prefer to maintain the old non-COARDS compliant format as their default grid format can instead select “cf”. Support for extracting 2-D slices from 3-D netCDF grids has also been added.
  2. An overhaul of how the pslib library encodes PostScript images, resulting in vastly smaller files when certain conditions are met, and general shrinking overall by enabling RLE or LZW compression. We have also added hooks for setting three new PostScript parameters via gmtdefaults settings: PS_LINE_CAP, PS_LINE_JOIN, and PS_MITER_LIMIT. See gmtdefaults for details.
  3. Improved alignment of strings ending in “1” in the PostScript output.
  4. Adjustments to how native GMT grid headers are read and written in order to be fully 64-bit safe. GMT now runs in full 64-bit mode on platforms that supports it (e.g., Mac OS X G5).
  5. Making GMT tread-safe by replacing strtok with our own GMT_strtok function.
  6. Implemented full inverse Winkel map projection based on a new algorithm by Ipbuker, 2002, Cartography & Geographical Information Science, 29, 37-42.
  7. Extended the options that is used to specify grid spacing (usually -Ixinc/yinc) to allow for specifying nx/ny instead (by appending +). Also, append ! to adjust the range so it fits exactly the given increment [by default the range is kept fixed and sloppy increments are adjusted accordingly]. Append e|k|i|n for increments in meter, km, miles or nautical miles, respectively. These increments are converted to degrees longitude (at the middle latitude) and degrees latitude.
  8. The polar r,θ projection -Jp now takes an optional suffix r that reverses the radial coordinates (useful when r is elevation as used by sky plots.)
  9. The misc supplement has two new items: ps2raster uses ghostscript to fascilitate the rasterization of PostScript files, while nc2xy allows extraction of data columns from COARDS-compliant netCDF files.
  10. The mgd77 supplement has two new items: mgd77convert translates between different MGD77 formats (including a new netCDF-based format), while mgd77manage assists in the management of trackline data sets.
  11. We now have improved PDF layout and navigation (thanks to Misha Tchernychev).
  12. The HTML versions of all manual pages are now generated with groff, and has active links for GMT Default parameters as they are references in the documentation.

Many programs or options have received minor updates and new features, such as

-b
: Ability to specify byte-swapping for native binary input and output tables by using upper case S|D. This is useful if you have binary tables created on a little-endian machine (e.g., Linux PC) and need to read them on a big-endian machine (e.g., most RISC-chip machines from Sun, HP, Apple).
filter1d
: Allow NaNs in all but the “independent data” column.
grdcontour
: Label option +ap[u|d] for always having labels be readable up or down hills.
gmtconvert
: New -N option suppresses output records when all fields are NaNs.
gmtmath
: Added TN function for evaluating Chebyshev polynomials; new constant Tn was added to easily select normalized T (gives coordinates from -1 to + 1 suitable for evaluating Legendre and Chebyshev polynomials). Finally, we added CORRCOEFF for calculation of correlation coefficients, and -I to reverse the output by printing the last row first.
grd2cpt
: New option -D sets the back- and foreground colors to the colors at the limits of the cpt file.
grd2xyz
: Added -E for ESRI interchange ASCII grid dump.
grdfilter
: Geographic boundary conditions are now in effect if -D4 is selected.
grdgradient
: Added option -E for Lambertian or Peuckeer illumination.
grdmath
: Allow -bi to be used with input files for commands PDIST, LDIST, and INSIDE. When spherical calculations are selected we now use the ELLIPSOID setting to determine if distance calculations should be along geodesics or great circles. Also added TN function for evaluating Chebyshev polynomials; new constants Xn and Yn was added to easily select normalized X and Y. Finally, we added CORRCOEFF for calculation of correlation coefficients.
grdraster
: Optionally select a data set by giving a text pattern instead of data ID number. This makes it easier to specify a certain data set (i.e., “ETOPO2”) than having to remember its arbitrary numerical ID. Also, native grids with GMT headers can also be placed in the database by appending Hnbytes to the corresponding grdraster.info entry, where nbytes is the size of the header that should be skipped (use 892 for GMT headers).
makecpt
: New option -D sets the back- and foreground colors to the colors at the limits of the cpt file.
mapproject
: -L now outputs both the minimum distance and the coordinates of the nearest point on the line.
pscoast
: Added -Z for 3-D map z-level (as in psbasemap and others).
pshistogram
: New option -Tcol lets user select any column to be used, starting with 0 (first). The old -2 option is retired (but remains accessible for backwards compatibility).
psimage
: Now support inclusion of EPS images.
pslegend
: Added layout option B for inserting color bars via psscale.
psscale
: Now supports an optional ;label at end of each line in cpt files. If present this label will replace the default annotations when option -L is used.
psxyz
: Added -Q to disable sorting of points based on distance.
sample1d
: Allow NaNs in all but the “independent data” column.
xyz2grd
: Added -E for ESRI interchange ASCII grid digest.

Inevitably, when new features are added, new bugs come along with them. Below is a list of problems that we have identified and corrected in the current release:

install_gmt
: No longer test netcdf installation since that can fail even when install was successful [e.g., under Mac OS X Tiger].
gmt.h
: GMT_swab4 used unsigned long instead of unsigned int which could cause 64-bit problems.
gmt_time_system.h
: Fixed MJD offsets by subtracting 10 days.
gmt_calclock.c
: time to hr,min,sec was vulnerable to round-off when optimized. Also, hh:mm data (without trailing :ss) would loose the minutes (hh:mm:ss was OK).
gmt_grdio.c
: Bug in scale/offset for grdblend’s row-by-row i/o.
gmt_init.c
: Would eat number with leading plus sign without checking if it actually was a +gmtdefaults file instruction; thus gmtmath could not see numbers such as +13.5. Command line argument –BASEMAP_FRAME_RGB=color was not passed through to tick-, grid- and annotation-properties. GMT_end now frees memory used for hashing. Did not use custom ellipsoid to set DEG2M parameter so we got large errors for planets with significantly different radii.
gmt_io.c
: Bug in reading yyyy[/]jjj data fixed. GMT_lines_init had trouble if 2000 segments had no data at all. It also allocated 2000 points for each segment but never deallocated the unused portions, thus running up memory fast. GMT_write_segmentheader wrote nothing if input was binary and output is ASCII. Fixed a few memory leaks.
gmt_map.c
: Azimuth to angle calculation for linear projections now properly handle different scales in x and y. The calculation was also vulnerable to bad wrap-around, giving strange directions for vectors in psxy. Geodesic distance calculation could get wrong quadrant.
gmt_plot.c
: 360° polar basemaps could lack outline. Direction for map roses were inaccurate. Circle and θ-r boundaries did not allocate enough memory for arrays. Would plot both -180 and +180 annotations for periodic maps.
gmt_shore.c
: Must explicitly close polygons since inside/outside test in other programs expects it.
gmt_support.c
: Trouble extracting subregions of grid with east = 0. Cartesian LDIST failed when mininum distance was requested (only done via grdmath). Color names got truncated to 16 characters. Improved workings of GMT_sample_cpt in support of makecpt. Fixed more memory leaks. Bad LF/CR removal for coastline.conf dir.
filter1d
: -Ff with even number of coefficients sometimes skip a coefficient.
gmtconvert
: Missed first multisegment output header if input file was ASCII.
gmtmath
: No longer have to say -Ca if there is only one input column. Did not understand dateTclock as command line data.
gmtselect
: If -M is on and a portion of a segment is skipped, we must reissue the multisegment header when segment resumes. Now handles both Cartesian and spherical polygons correctly.
grd2xyz
: Sloppy -R resulted in bad x,y values and sometimes allocation error.
grdfilter
: Convolution filters now use correct area normalization.
grdgradient
: If -M is used with grids that include poles, ignore the poles when normalizing the slopes.
grdimage
: Cannot use -R to extract subset when -J is oblique. Reverse log-axes did not work.
grdmask
: Now handles both Cartesian and spherical correctly.
grdmath
: Wrong sign in D2DY2, and bogus value at y = ymin. Now handles both Cartesian and spherical polygons correctly. Constants given on command line can now be absolute time, geographic coordinates, or regular floating-point numbers.
grdtrack
: Would fail to skip first two columns for ASCII input if dd:mm:ss format was used.
grdview
: Cannot use -R to extract subset when -J is oblique.
grdvolume
: -Clow/high/delta did not check for low < high, etc.
pscoast
: Recursive painting could get tricked when boundaries were curved.
pslegend
: Did not pass +gmtdefaults and –PAR=val onto system calls.
psscale
: Vertical annotations w/custom D_FORMAT were not aligned. Now uses more optimal means to display the color bar, leading to smaller PostScript files. -E did not flip sides when a negative width was used.
psxy
: -Sp is now a true point, but can also take an optional size. Pentagon symbol had wrong normalization scale. If a fixed symbol size was given in -S, with the symbol type supplied from file, we would not scale symbols correctly if upper case symbols were given.
psxyz
: Wrong index used in assigning color from cpt and in updating vector lengths. If a fixed symbol size was given in -S, with the symbol type supplied from file, we would not scale symbols correctly if upper case symbols were given
spectrum1d
: Bugs in error expressions for admittance, gain, and phase have been corrected.
x2sys & mgd77 supplements
: Made DOS-format (CR/LF) tolerant. Both supplements are now undergoing rapid development.

1.1.29 Overview of GMT 4.0 [Oct-10, 2004]

GMT 4 represents a major overhaul of the package, hence the major version number increment. There are four categories of changes that have been implemented:

Time-series support.
GMT can now read and write time-series data where the time coordinates are of the form dateTclock2. The formats used for date and clock are under the user’s control. Both Gregorian and ISO calendars are supported. Frame annotation for time-series are now supported via the -B option; there are many new modifiers reflecting the vast number of ways one may want to annotate time axes, including support for primary and secondary annotation levels and the day- and month-names in numerous languages (send us the information we need if your language is not supported). The capability to handle time (in -R, -J, -B, i/o, and plotting) required considerable changes “under the hood”, including the introduction of numerous new gmtdefaults parameters to make the time series support as “generic” as we need it to be.
New Tools.
Three new tools have been added:
  1. gmt2rgb: Makes red, green, and blue component grid files from an image (to be used with new options for false color imaging or image draping by grdimage or grdview).
  2. grdblend: Blends several partially over-lapping grid files into one combined grid. Output grid is written one row at the time so truly enormous grids can be created.
  3. pslegend: Designs and plots elaborate legends on maps.
New Program Options.
Many programs have received additional options or features that enhances their usefulness:
General enhancements.
These affect most of the programs:

Chapter 2
Introduction

Most scientists are familiar with the sequence: raw data processing final illustration. In order to finalize papers for submission to scientific journals, prepare proposals, and create overheads and slides for various presentations, many scientists spend large amounts of time and money to create camera-ready figures. This process can be tedious and is often done manually, since available commercial or in-house software usually can do only part of the job. To expedite this process we introduce the Generic Mapping Tools (GMT for short), which is a free4, software package that can be used to manipulate columns of tabular data, time-series, and gridded data sets, and display these data in a variety of forms ranging from simple x-y plots to maps and color, perspective, and shaded-relief illustrations. GMT uses the PostScript page description language [Adobe Systems Inc., 1990]. With PostScript, multiple plot files can easily be superimposed to create arbitrarily complex images in gray tones or 24-bit true color. Line drawings, bitmapped images, and text can be easily combined in one illustration. PostScript plot files are device-independent: The same file can be printed at 300 dots per inch (dpi) on an ordinary laserwriter or at 2470 dpi on a phototypesetter when ultimate quality is needed. GMT software is written as a set of UNIX tools5 and is totally self-contained and fully documented. The system is offered free of charge and is distributed over the computer network (Internet) [Wessel and Smith, 1991; 1995a,b; 1998].

The original version 1.0 of GMT was released in the summer of 1988 when the authors were graduate students at Lamont-Doherty Earth Observatory of Columbia University. During our tenure as graduate students, L-DEO changed its computing environment to a distributed network of UNIX workstations, and we wrote GMT to run in this environment. It became a success at L-DEO, and soon spread to numerous other institutions in the US, Canada, Europe, and Japan. The current version benefits from the many suggestions contributed by users of the earlier versions, and now includes more than 50 tools, more than 30 projections, and many other new, more flexible features. GMT provides scientists with a variety of tools for data manipulation and display, including routines to sample, filter, compute spectral estimates, and determine trends in time series, grid or triangulate arbitrarily spaced data, perform mathematical operations (including filtering) on 2-D data sets both in the space and frequency domain, sample surfaces along arbitrary tracks or onto a new grid, calculate volumes, and find trend surfaces. The plotting programs will let the user make linear, log10, and xayb diagrams, polar and rectangular histograms, maps with filled continents and coastlines choosing from many common map projections, contour plots, mesh plots, monochrome or color images, and artificially illuminated shaded-relief and 3-D perspective illustrations.

GMT is written in the highly portable ANSI C programming language [Kernighan and Ritchie, 1988], is fully POSIX compliant [Lewine, 1991], has no Year 2000 problems, and may be used with any hardware running some flavor of UNIX, possibly with minor modifications. In writing GMT, we have followed the modular design philosophy of UNIX: The raw data processing final illustration flow is broken down to a series of elementary steps; each step is accomplished by a separate GMT or UNIX tool. This modular approach brings several benefits: (1) only a few programs are needed, (2) each program is small and easy to update and maintain, (3) each step is independent of the previous step and the data type and can therefore be used in a variety of applications, and (4) the programs can be chained together in shell scripts or with pipes, thereby creating a process tailored to do a user-specific task. The decoupling of the data retrieval step from the subsequent massage and plotting is particularly important, since each institution will typically have its own data base formats. To use GMT with custom data bases, one has only to write a data extraction tool which will put out data in a form readable by GMT (discussed below). After writing the extractor, all other GMT modules will work as they are.

GMT makes full use of the PostScript page description language, and can produce color illustrations if a color PostScript device is available. One does not necessarily have to have access to a top-of-the-line color printer to take advantage of the color capabilities offered by GMT: Several companies offer imaging services where the customer provides a PostScript plot file and gets color slides or hardcopies in return. Furthermore, general-purpose PostScript raster image processors (RIPs) are now becoming available, letting the user create raster images from PostScript and plot these bitmaps on raster devices like computer screens, dot-matrix printers, large format raster plotters, and film writers6. Because the publication costs of color illustrations are high, GMT offers 90 common bit and hachure patterns, including many geologic map symbol types, as well as complete graytone shading operations. Additional bit and hachure patterns may also be designed by the user. With these tools, it is possible to generate publication-ready monochrome originals on a common laserwriter.

GMT is thoroughly documented and comes with a technical reference and cookbook which explains the purpose of the package and its many features, and provides numerous examples to help new users quickly become familiar with the operation and philosophy of the system. The cookbook contains the shell scripts that were used for each example; PostScript files of each illustration are also provided. All programs have individual manual pages which can be installed as part of the on-line documentation under the UNIX man utility or as web pages. In addition, the programs offer friendly help messages which make them essentially self-teaching – if a user enters invalid or ambiguous command arguments, the program will print a warning to the screen with a synopsis of the valid arguments. All the documentation is available for web browsing and may be installed at the user’s site.

The processing and display routines within GMT are completely general and will handle any (x,y) or (x,y,z) data as input. For many purposes the (x,y) coordinates will be (longitude, latitude) but in most cases they could equally well be any other variables (e.g., wavelength, power spectral density). Since the GMT plot tools will map these (x,y) coordinates to positions on a plot or map using a variety of transformations (linear, log-log, and several map projections), they can be used with any data that are given by two or three coordinates. In order to simplify and standardize input and output, GMT uses two file formats only. Arbitrary sequences of (x,y) or (x,y,z) data are read from multi-column ASCII tables, i.e., each file consists of several records, in which each coordinate is confined to a separate column7. This format is straightforward and allows the user to perform almost any simple (or complicated) reformatting or processing task using standard UNIX utilities such as cut, paste, grep, sed and awk. Two-dimensional data that have been sampled on an equidistant grid are read and written by GMT in a binary grid file using the functions provided with the netCDF library (a free, public-domain software library available separately from UCAR, the University Corporation of Atmospheric Research [Treinish and Gough, 1987]). This XDR (External Data Representation) based format is architecture independent, which allows the user to transfer the binary data files from one computer system to another8. GMT contains programs that will read ASCII (x,y,z) files and produce grid files. One such program, surface, includes new modifications to the gridding algorithm developed by Smith and Wessel [1990] using continuous splines in tension.

Most of the programs will produce some form of output, which falls into four categories. Several of the programs may produce more than one of these types of output:

  1. 1-D ASCII Tables — For example, a (x,y) series may be filtered and the filtered values output. ASCII output is written to the standard output stream.
  2. 2-D binary (netCDF or user-defined) grid files – Programs that grid ASCII (x,y,z) data or operate on existing grid files produce this type of output.
  3. PostScript – The plotting programs all use the PostScript page description language to define plots. These commands are stored as ASCII text and can be edited should you want to customize the plot beyond the options available in the programs themselves.
  4. Reports – Several GMT programs read input files and report statistics and other information. Nearly all programs have an optional “verbose” operation, which reports on the progress of computation. All programs feature usage messages, which prompt the user if incorrect commands have been given. Such text is written to the standard error stream and can therefore be separated from ASCII table output.

GMT is available over the Internet at no charge. To obtain a copy, visit the GMT home page gmt.soest.hawaii.edu, and select DOWNLOAD. This page contains all the information you need to obtain GMT for your platform. We also maintain two electronic mailing lists you may subscribe to in order to stay informed about bug fixes and upgrades (See Chapter 7).

For those without net-access that need to obtain GMT: Geoware (http://www.geoware-online.com) makes and distributes CD-R and DVD-R media with the GMT package, compatible supplements, and several Gb of useful Earth and ocean science data sets. For more information send e-mail to geoware@geoware-online.com.

GMT has served a multitude of scientists very well, and their responses have prompted us to develop these programs even further. It is our hope that the new version will satisfy these users and attract new users as well. We present this system to the community in order to promote sharing of research software among investigators in the US and abroad.

References

  1. Kernighan, B. W., and D. M. Ritchie, The C programming language, 2nd edition, p. 272, Prentice-Hall, Englewood Cliffs, New Jersey, 1988.
  2. Adobe Systems Inc., PostScript Language Reference Manual, 2nd edition, p. 764, Addison-Wesley, Reading, Massachusetts, 1990.
  3. Lewine, D., POSIX programmer’s guide, 1st edition, p. 607, O’Reilly & Associates, Sebastopol, California, 1991.
  4. Treinish, L. A., and M. L. Gough, A software package for the data-independent management of multidimensional data, EOS trans. AGU, 68, 633–635, 1987.
  5. Smith, W. H. F., and P. Wessel, Gridding with continuous curvature splines in tension, Geophysics, 55, 293–305, 1990.
  6. Wessel, P., and W. H. F. Smith, New, improved version of Generic Mapping Tools released, EOS trans. AGU, 79, 579, 1998.
  7. Wessel, P., and W. H. F. Smith, New version of the Generic Mapping Tools released, EOS trans. AGU, 76, 329, 1995a.
  8. Wessel, P., and W. H. F. Smith, New version of the Generic Mapping Tools released, EOS electronic supplement, http://www.agu.org/eos_elec/95154e.html, 1995b.
  9. Wessel, P., and W. H. F. Smith, Free software helps map and display data, EOS trans. AGU, 72, 441 & 445–446, 1991.

Chapter 3
GMT overview and quick reference

3.1 GMT summary

The following is a summary of all the programs supplied with GMT and a very short description of their purpose. For more details, see the individual UNIX manual pages or the online web documentation. For a listing sorted by program purpose, see Section 3.2.

blockmean L2 (x,y,z) table data filter/decimator

blockmedian L1 (x,y,z) table data filter/decimator

blockmode Mode estimate (x,y,z) table data filter/decimator

filter1d Filter 1-D table data sets (time series)

fitcircle Finds the best-fitting great or small circle for a set of points

gmt2rgb Convert Sun raster or grid file to red, green, blue component grids

gmtconvert Convert data tables from one format to another

gmtdefaults List the current default settings

gmtmath Mathematical operations on table data

gmtselect Select subsets of table data based on multiple spatial criteria

gmtset Change selected parameters in current .gmtdefaults4 file

grd2cpt Make color palette table from a grid files

grd2xyz Conversion from 2-D grid file to table data

grdblend Blend several partially over-lapping grid files onto one grid

grdclip Limit the z-range in gridded data sets

grdcontour Contouring of 2-D gridded data sets

grdcut Cut a sub-region from a grid file

grdedit Modify header information in a 2-D grid file

grdfft Perform operations on grid files in the frequency domain

grdfilter Filter 2-D gridded data sets in the space domain

grdgradient Compute directional gradient from grid files

grdhisteq Histogram equalization for grid files

grdimage Produce images from 2-D gridded data sets

grdinfo Get information about grid files

grdlandmask Create masking grid files from shoreline data base

grdmask Reset grid nodes in/outside a clip path to constants

grdmath Mathematical operations on grid files

grdpaste Paste together grid files along a common edge

grdproject Project gridded data sets onto a new coordinate system

grdreformat Converts grid files into other grid formats

grdsample Resample a 2-D gridded data set onto a new grid

grdtrack Sampling of 2-D gridded data set along 1-D track

grdtrend Fits polynomial trends to grid files

grdvector Plotting of 2-D gridded vector fields

grdview 3-D perspective imaging of 2-D gridded data sets

grdvolume Calculate volumes under a surface within specified contour

greenspline Interpolation using Green’s functions for splines in 1–3 dimensions

makecpt Make color palette tables

mapproject Transformation of coordinate systems for table data

minmax Report extreme values in table data files

nearneighbor Nearest-neighbor gridding scheme

project Project table data onto lines or great circles

ps2raster Crop and convert PostScript files to raster images, EPS, and PDF

psbasemap Create a basemap plot

psclip Use polygon files to define clipping paths

pscoast Plot (and fill) coastlines, borders, and rivers on maps

pscontour Contour or image raw table data by triangulation

pshistogram Plot a histogram

psimage Plot Sun raster files on a map

pslegend Plot a legend on a map

psmask Create overlay to mask out regions on maps

psrose Plot sector or rose diagrams

psscale Plot gray scale or color scale on maps

pstext Plot text strings on maps

pswiggle Draw table data time-series along track on maps

psxy Plot symbols, polygons, and lines on maps

psxyz Plot symbols, polygons, and lines in 3-D

sample1d Resampling of 1-D table data sets

spectrum1d Compute various spectral estimates from time-series

splitxyz Split xyz files into several segments

surface A continuous curvature gridding algorithm

trend1d Fits polynomial or Fourier trends to y = f(x) series

trend2d Fits polynomial trends to z = f(x,y) series

triangulate Perform optimal Delauney triangulation and gridding

xyz2grd Convert an equidistant table xyz file to a 2-D grid file

3.2 GMT quick reference

Instead of an alphabetical listing, this section contains a summary sorted by program purpose. Also included is a quick summary of the standard command line options and a breakdown of the -J option for each of the over 30 projections available in GMT.

FILTERING OF 1-D AND 2-D DATA


blockmean L 2 estimate (x,y,z) data filters/decimators


blockmedianL1 estimate (x,y,z) data filters/decimators


blockmode Mode estimate (x,y,z) data filters/decimators


filter1d Filter 1-D data (time series)


grdfilter Filter 2-D data in space domain


PLOTTING OF 1-D and 2-D DATA


grdcontour Contouring of 2-D gridded data


grdimage Produce images from 2-D gridded data


grdvector Plot vector fields from 2-D gridded data


grdview 3-D perspective imaging of 2-D gridded data


psbasemap Create a basemap frame


psclip Use polygon files as clipping paths


pscoast Plot coastlines, filled continents, rivers, and political borders


pscontour Direct contouring or imaging of xyz data by triangulation


pshistogramPlot a histogram


psimage Plot Sun raster files on a map


pslegend Plot a legend on a map


psmask Create overlay to mask specified regions of a map


psrose Plot sector or rose diagrams


psscale Plot gray scale or color scale


pstext Plot text strings


pswiggle Draw anomalies along track


psxy Plot symbols, polygons, and lines in 2-D


psxyz Plot symbols, polygons, and lines in 3-D


GRIDDING OF (X,Y,Z) TABLE DATA


greenspline Interpolation using Green’s functions for splines in 1–3 dimensions


nearneighborNearest-neighbor gridding scheme


surface Continuous curvature gridding algorithm


triangulate Perform optimal Delauney triangulation on xyz data


SAMPLING OF 1-D AND 2-D DATA


grdsample Resample a 2-D gridded data onto new grid


grdtrack Sampling of 2-D data along 1-D track


sample1d Resampling of 1-D data


PROJECTION AND MAP-TRANSFORMATION


grdproject Transform gridded data to a new coordinate system


mapproject Transform table data to a new coordinate system


project Project data onto lines or great circles


INFORMATION


gmtdefaults List the current default settings


gmtset Command-line editing of parameters in the .gmtdefaults4 file


grdinfo Get information about the content of grid files


minmax Report extreme values in table data files


MISCELLANEOUS


gmtmath Reverse Polish Notation (RPN) calculator for table data


makecpt Create GMT color palette tables


spectrum1d Compute spectral estimates from time-series


triangulate Perform optimal Delauney triangulation on xyz data


CONVERT OR EXTRACT SUBSETS OF DATA


gmt2rgb Convert Sun raster or grid file to red, green, blue component grids


gmtconvert Convert table data from one format to another


gmtselect Select table data subsets based on multiple spatial criteria


grd2xyz Convert 2-D gridded data to table data


grdcut Cut a sub-region from a grid file


grdblend Blend several partially over-lapping grid files onto one grid


grdpaste Paste together grid files along common edge


grdreformat Convert from one grid format to another


splitxyz Split (x,y,z) table data into several segments


xyz2grd Convert table data to 2-D grid file


DETERMINE TRENDS IN 1-D AND 2-D DATA


fitcircle Finds best-fitting great or small circles


grdtrend Fits polynomial trends to grid files (z = f(x,y))


trend1d Fits polynomial or Fourier trends to y = f(x) series


trend2d Fits polynomial trends to z = f(x,y) series


OTHER OPERATIONS ON 2-D GRIDS


grd2cpt Make color palette table from grid file


grdclip Limit the z–range in gridded data sets


grdedit Modify grid header information


grdfft Operate on grid files in frequency domain


grdgradient Compute directional gradients from grid files


grdhisteq Histogram equalization for grid files


grdlandmask Creates mask grid file from coastline database


grdmask Set grid nodes in/outside a clip path to constants


grdmath Reverse Polish Notation (RPN) calculator for grid files


grdvolume Calculate volume under a surface within a contour


MANIPULATING GMT POSTSCRIPT FILES


ps2raster Crop and convert PostScript files to raster images, EPS and PDF


STANDARDIZED COMMAND LINE OPTIONS WITH OLD PROJECTION CODES


-B[p|s]xinfo[/yinfo[/zinfo]][WESNZwesnz+][:.title:] Tickmarks. Each info is
  [t]stride[±phase][u][l|p][:"label":][:="prefix":][:,"unit":], where l and p apply to log10 axes only,
  and type t = {a, A, f, g, i, I}; unit u = {c, C, d, D, h, H, K, k, m, M, o, O, r, R, u, U, y, Y}
  The leading p|s selects primary [Default] or secondary axis items


-H[i][n_headers] ASCII [input] tables have header record[s]


-J (upper case for width, lower case for scale) Map projection


  -JAlon0/lat0[/horizon]/width Lambert azimuthal equal area


  -JBlon0/lat0/lat1/lat2/width Albers conic equal area


  -JClon0/lat0/width Cassini cylindrical


  -JCyl_stere/[lon0/[lat0/]]width Cylindrical stereographic


  -JDlon0/lat0/lat1/lat2/width Equidistant conic


  -JElon0/lat0[/horizon]/width Azimuthal equidistant


  -JFlon0/lat0[/horizon]/width Azimuthal gnomonic


  -JGlon0/lat0[/horizon]/width Azimuthal orthographic


  -JGlon0/lat0/alt/azim/tilt/twist/W/H/width General perspective


  -JH[lon0/]width Hammer equal area


  -JI[lon0/]width Sinusoidal equal area


  -JJ[lon0/]width Miller cylindrical


  -JKf[lon0/]width Eckert IV equal area


  -JKs[lon0/]width Eckert VI equal area


  -JLlon0/lat0/lat1/lat2/width Lambert conic conformal


  -JM[lon0/[lat0/]]width Mercator cylindrical


  -JN[lon0/]width Robinson


  -JOalon0/lat0/azim/width Oblique Mercator, 1: origin and azimuth


  -JOblon0/lat0/lon1/lat1/width Oblique Mercator, 2: two points


  -JOclon0/lat0/lonp/latp/width Oblique Mercator, 3: origin and pole


  -JP[a]width[/origin] Polar [azimuthal] (θ,r) (or cylindrical)


  -JPoly[lon0/[lat0/]]width (American) polyconic


  -JQ[lon0/[lat0/]]width Equidistant cylindrical


  -JR[lon0/]width Winkel Tripel


  -JSlon0/lat0/[/horizon]/width General stereographic


  -JTlon0/[lat0/]width Transverse Mercator


  -JUzone/width Universal Transverse Mercator (UTM)


  -JV[lon0/]width Van der Grinten


  -JW[lon0/]width Mollweide


  -JXwidth[l|pexp|T|t][/height[l|pexp|T|t]][d]Linear, log10, xayb, and time


  -JYlon0/lat0/width Cylindrical equal area


-K Append more PS later


-O This is an overlay plot


-P Select Portrait orientation


-Rwest/east/south/north[/zmin/zmax][r] Specify Region of interest


-U[[just]/dx/dy/][label] Plot time-stamp on plot


-V Run in verbose mode


-X[a|c|r]off [u] Shift plot origin in x-direction


-Y[a|c|r]off [u] Shift plot origin in y-direction


-b[i|o][c|s|S|d|D][ncol] Select binary input or output


-ccopies Set number of plot copies [1]


-f[i|o]colinfo Set formatting of ASCII input or output


-g[+]x|X|y|Y|d|Dgap[u] Segment data by detecting gaps


-m[i|o]flag Set multi-segment data mode


-:[i|o] Expect y/x input rather than x/y


STANDARDIZED COMMAND LINE OPTIONS (WITH Proj4 PROJECTION CODES)


-B[p|s]xinfo[/yinfo[/zinfo]][WESNZwesnz+][:.title:] Tickmarks. Each info is
  [t]stride[±phase][u][l|p][:"label":][:="prefix":][:,"unit":], where l and p apply to log10 axes only,
  and type t = {a, A, f, g, i, I}; unit u = {c, C, d, D, h, H, K, k, m, M, o, O, r, R, u, U, y, Y}
  The leading p|s selects primary [Default] or secondary axis items


-H[i][n_headers] ASCII [input] tables have header record[s]


-J (upper case for width, lower case for scale) Map projection


  -Jaea/lon0/lat0/lat1/lat2/scale Albers conic equal area


  -Jaeqd/lon0/lat0[/horizon]/scale Azimuthal equidistant


  -Jcass/lon0/lat0/scale Cassini cylindrical


  -Jcea/lon0/lat0/scale Cylindrical equal area


  -Jcyl_stere/[lon0/[lat0/]]scale Cylindrical stereographic


  -Jeqc/[lon0/[lat0/]]scale Equidistant cylindrical


  -Jeqdc/lon0/lat0/lat1/lat2/scale Equidistant conic


  -Jgnom/lon0/lat0[/horizon]/scale Azimuthal gnomonic


  -Jhammer/[lon0/]scale Hammer equal area


  -Jeck4/[lon0/]scale Eckert IV equal area


  -Jeck6/[lon0/]scale Eckert VI equal area


  -Jlaea/lon0/lat0[/horizon]/scale Lambert azimuthal equal area


  -Jlcc/lon0/lat0/lat1/lat2/scale Lambert conic conformal


  -Jmerc/[lon0/[lat0/]]scale Mercator cylindrical


  -Jmill/[lon0/]scale Miller cylindrical


  -Jmoll/[lon0/]scale Mollweide


  -Jnsper/lon0/lat0/alt/azim/tilt/twist/W/H/scale General perspective


  -Jomerc/lon0/lat0/azim/scale Oblique Mercator, 1: origin and azimuth


  -Jomerc/lon0/lat0/lon1/lat1/scale Oblique Mercator, 2: two points


  -Jomercp/lon0/lat0/lonp/latp/scale Oblique Mercator, 3: origin and pole


  -Jortho/lon0/lat0[/horizon]/scale Azimuthal orthographic


  -Jpolar/[a]scale[/origin] Polar [azimuthal] (θ,r) (or cylindrical)


  -Jpoly[lon0/[lat0/]]width (American) polyconic


  -Jrobin/[lon0/]scale Robinson


  -Jsinu/[lon0/]scale Sinusoidal equal area


  -Jstere/lon0/lat0/[/horizon]/scale General stereographic


  -Jtmerc/lon0/[lat0/]scale Transverse Mercator


  -Jutm/zone/scale Universal Transverse Mercator (UTM)


  -Jvandg/[lon0/]scale Van der Grinten


  -Jwintri/[lon0/]scale Winkel Tripel


  -Jxy/xscale[l|pexp|T|t][/yscale[l|pexp|T|t]][d]Linear, log10, xayb, and time


-K Append more PS later


-O This is an overlay plot


-P Select Portrait orientation


-Rwest/east/south/north[/zmin/zmax][r] Specify Region of interest


-U[[just]/dx/dy/][label] Plot time-stamp on plot


-V Run in verbose mode


-X[a|c|r]off [u] Shift plot origin in x-direction


-Y[a|c|r]off [u] Shift plot origin in y-direction


-b[i|o][c|s|S|d|D][ncol] Select binary input or output


-ccopies Set number of plot copies [1]


-f[i|o]colinfo Set formatting of ASCII input or output


-g[+]x|X|y|Y|d|Dgap[u] Segment data by detecting gaps


-m[i|o]flag Set multi-segment data mode


-:[i|o] Expect y/x input rather than x/y


Chapter 4
General features

This section explains features common to all the programs in GMT and summarizes the philosophy behind the system. Some of the features described here may make more sense once you reach the cook-book section where we present actual examples of their use.

4.1 GMT units

GMT programs can accept dimensional quantities in cm, inch, meter, or point (1/72 of an inch)9. There are two ways to ensure that GMT understands which unit you intend to use.

  1. Append the desired unit to the dimension you supply. This way is explicit and clearly communicates what you intend, e.g., -X4c means the length being passed to the -X switch is 4 cm.
  2. Set the parameter MEASURE_UNIT to the desired unit. Then, all dimensions without explicit unit will be interpreted accordingly.

The latter method is less secure as other users may have a different unit set and your script may not work as intended. We therefore recommend you always supply the desired unit explicitly.

4.2 GMT defaults

4.2.1 Overview and the .gmtdefaults4 file


PIC

Figure 4.1: Some GMT parameters that affect plot appearance.

There are about 100 parameters which can be adjusted individually to modify the appearance of plots or affect the manipulation of data. When a program is run, it initializes all parameters to the GMT defaults10, then tries to open the file .gmtdefaults4 in the current directory11. If not found, it will look for that file in a sub-directory /̃.gmt of your home directory, and finally in your home directory itself. If successful, the program will read the contents and set the default values to those provided in the file. By editing this file you can affect features such as pen thicknesses used for maps, fonts and font sizes used for annotations and labels, color of the pens, dots-per-inch resolution of the hardcopy device, what type of spline interpolant to use, and many other choices (A complete list of all the parameters and their default values can be found in the gmtdefaults manual pages). Figures 4.2.1, 4.2, and 4.3 show the parameters that affect plots). You may create your own .gmtdefaults4 files by running gmtdefaults and then modify those parameters you want to change. If you want to use the parameter settings in another file you can do so by specifying +<defaultfile> on the command line. This makes it easy to maintain several distinct parameter settings, corresponding perhaps to the unique styles required by different journals or simply reflecting font changes necessary to make readable overheads and slides. Note that any arguments given on the command line (see below) will take precedent over the default values. E.g., if your .gmtdefaults4 file has x offset = 1i as default, the -X1.5i option will override the default and set the offset to 1.5 inches.

There are at least two good reasons why the GMT default options are placed in a separate parameter file:

  1. It would not be practical to allow for command-line syntax covering so many options, many of which are rarely or never changed (such as the ellipsoid used for map projections).
  2. It is convenient to keep separate .gmtdefaults4 files for specific projects, so that one may achieve a special effect simply by running GMT commands in the directory whose .gmtdefaults4 file has the desired settings. For example, when making final illustrations for a journal article one must often standardize on font sizes and font types, etc. Keeping all those settings in a separate .gmtdefaults4 file simplifies this process and will allow you to generate those illustrations with the same settings later on. Likewise, GMT scripts that make figures for PowerPoint presentations often use a different color scheme and font size than output intended for laser printers. Organizing these various scenarios into separate .gmtdefaults4 files will minimize headaches associated with micro-editing of illustrations.
    PIC
    Figure 4.2: More GMT parameters that affect plot appearance.

4.2.2 Changing GMT defaults

As mentioned, GMT programs will attempt to open a file named .gmtdefaults4. At times it may be desirable to override that default. There are several ways in which this can be accomplished.

  1. Supply another filename using the +filename syntax, i.e., on the same command line as the GMT command we append the name of the alternate .gmtdefaults4 file with the plus sign as a prefix. Because any changes only apply to that one command you would have to append the alternate file to every command in your script. This is tedious but may be an option for situations when you cannot write in the current directory (e.g., some CGI scripts).
  2. A perhaps less tedious method is to start each script by making a copy of the current .gmtdefaults4, then copy the desired .gmtdefaults4 file to the current directory, and finally undo the changes at the end of the script. Possible side effects include premature ending of the script due to user error or bugs which means the final resetting does not take place (unless you write your script very carefully.)
  3. To permanently change some of the GMT parameters on the fly inside a script the gmtset utility can be used. E.g., to change the primary annotation font to 12 point Times-Bold we run

    gmtset ANNOT_FONT_PRIMARY Times-Bold ANNOT_FONT_SIZE_PRIMARY 12

    These changes will remain in effect until they are overridden.

  4. If all you want to achieve is to change a few parameters during the execution of a single command but otherwise leave the environment intact, consider passing the parameter changes on the command line via the ––PAR=value mechanism. For instance, to temporarily set the output format for floating points to have lots of decimals, say, for map projection coordinate output, append ––D_FORMAT=%.12lg to the command in question.
  5. Finally, since version 4.2.2 GMT provides to possibility to override the settings only during the running of a single script, reverting to the original settings after the script is run, as if the script was run in “isolation”. The isolation mode is discussed in Section P.1.

In addition to those parameters that directly affect the plot there are numerous parameters than modify units, scales, etc. For a complete listing, see the gmtdefaults man pages. We suggest that you go through all the available parameters at least once so that you know what is available to change via one of the described mechanisms.


PIC

Figure 4.3: Even more GMT parameters that affect plot appearance.

Note: All examples presented in this document started by copying the file .gmtdefaults4.doc from the directory doc/scripts to .gmtdefaults4. As a result the commands gmtset of other scripts were overall, reverting to a “virgin” of parameters set in .gmtdefaults4.doc. The graphs in Chapter 7 were created using .gmtdefaults4.doc from the directory examples after which the graphs were scaled down by 50%.

4.3 Command line arguments

Each program requires certain arguments specific to its operation. These are explained in the manual pages and in the usage messages. Most programs are “case-sensitive”; almost all options must start with an upper-case letter. We have tried to choose letters of the alphabet which stand for the argument so that they will be easy to remember. Each argument specification begins with a hyphen (except input file names; see below), followed by a letter, and sometimes a number or character string immediately after the letter. Do not space between the hyphen, letter, and number or string. Do space between options. Example:

pscoast -R0/20/0/20 -Ggray -JM6i -Wthin -B5 -V -P > map.ps

4.4 Standardized command line options

Most of the programs take many of the same arguments like those related to setting the data region, the map projection, etc. The 17 switches in Table 4.1 have the same meaning in all the programs (although some programs may not use all of them). These options will be described here as well as in the manual pages, as is vital that you understand how to use these options. We will present these options in order of importance (some are use a lot more than others).




Option
Meaning


-B Defines tickmarks, annotations, and labels for basemaps and axes


-H Specifies that input/output tables have header record(s)


-J Selects a map projection or coordinate transformation


-K Allows more plot code to be appended to this plot later


-O Allows this plot code to be appended to an existing plot


-P Selects Portrait plot orientation [Default is landscape]


-R Defines the extent of the map/plot region


-U Plots a time-stamp, by default in the lower left corner of page


-V Selects verbose operation; reporting on progress


-X Sets the x-coordinate for the plot origin on the page


-Y Sets the y-coordinate for the plot origin on the page


-b Selects binary input and/or output


-c Specifies the number of plot copies


-f Specifies the data format on a per column basis


-g Identify data gaps based on supplied criteria


-m Specifies data in multiple segment format


-: Assumes input geographic data are (lat,lon) and not (lon,lat)



Table 4.1: The 17 standardized GMT command line switches.

4.4.1 Data domain or map region: The -R option


PIC

Figure 4.4: The plot region can be specified in two different ways. (a) Extreme values for each dimension, or (b) coordinates of lower left and upper right corners.

The -R option defines the map region or data domain of interest. It may be specified in one of three ways (Figure 4.4):

  1. -Rxmin/xmax/ymin/ymax. This is the standard way to specify Cartesian data domains and geographical regions when using map projections where meridians and parallels are rectilinear.
  2. -Rxlleft/ylleft/xuright/yurightr. This form is used with map projections that are oblique, making meridians and parallels poor choices for map boundaries. Here, we instead specify the lower left corner and upper right corner geographic coordinates, followed by the suffix r.
  3. -Rgridfile. This will copy the domain settings found for the grid in specified file. Note that depending on the nature of the calling program, this mechanism will also set grid spacing and possibly the grid registration (see Appendix B.2.2).

For rectilinear projections the first two forms give identical results. Depending on the selected map projection (or the kind of expected input data), the boundary coordinates may take on three different formats:

Geographic coordinates:
These are longitudes and latitudes and may be given in decimal degrees (e.g., -123.45417) or in the [±]ddd[:mm[:ss[.xxx]]][W|E|S|N] format (e.g., 123:27:15W). Note that -Rg and -Rd are shorthands for “global domain” -R0/360/-90/90 and -R-180/180/-90/90, respectively.

When used in conjunction with the Cartesian Linear Transformation (-Jx or -JX) —which can be used to map floating point data, geographical coordinates, as well as time coordinates— it is prudent to indicate that you are using geographical coordinates in one of the following ways:

Calendar time coordinates:
These are absolute time coordinates referring to a Gregorian or ISO calendar. The general format is [date]T[clock], where date must be in the yyyy[-mm[-dd]] (year, month, day-of-month) or yyyy[-jjj] (year and day-of-year) for Gregorian calendars and yyyy[-Www[-d]] (year, week, and day-of-week) for the ISO calendar. If no date is given we assume the current day. Following the [optional] date string we require the T flag.

The optional clock string is a 24-hour clock in hh[:mm[:ss[.xxx]]] format. If no clock is given it implies 00:00:00, i.e., the start of the specified day. Note that not all of the specified entities need be present in the data. All calendar date-clock strings are internally represented as double precision seconds since proleptic Gregorian date Mon Jan 1 00:00:00 0001. Proleptic means we assume that the modern calendar can be extrapolated forward and backward; a year zero is used, and Gregory’s reforms12 are extrapolated backward. Note that this is not historical.

Relative time coordinates:
These are coordinates which count seconds, hours, days or years relative to a given epoch. A combination of the parameters TIME_EPOCH and TIME_UNIT define the epoch and time unit. The parameter TIME_SYSTEM provides a few shorthands for common combinations of epoch and unit, like j2000 for days since noon of 1 Jan 2000. Denote relative time coordinates by appending the optional lower case t after the value. When it is otherwise apparent that the coordinate is relative time (for example by using the -f switch), the t can be omitted.
Other coordinates:
These are simply any coordinates that are not related to geographic or calendar time or relative time and are expected to be simple floating point values such as [±]xxx.xxx[E|e|D|d[±]xx], i.e., regular or exponential notations, with the enhancement to understand FORTRAN double precision output which may use D instead of E for exponents. These values are simply converted as they are to internal representation.13

4.4.2 Coordinate transformations and map projections: The -J option

This option selects the coordinate transformation or map projection. The general format is

Since GMT version 4.3.0, there is an alternative way to specify the projections: use the same abbreviation as in the mapping package Proj4. The options thus either look like:


PIC

Figure 4.5: The 30+ map projections and coordinate transformations available in GMT.

The projections available in GMT are presented in Figure 4.5. For details on all GMT projections and the required parameters, see the psbasemap man page. We will also show examples of every projection in the next Chapters, and a quick summary of projection syntax was given in Chapter 3.

4.4.3 Map frame and axes annotations: The -B option

This is by far the most complicated option in GMT, but most examples of its usage are actually quite simple. Given as -B[p|s]xinfo[/yinfo[/zinfo]][:."title string":][W|w][E|e][S|s][N|n][Z|z[+]], this switch specifies map boundaries (or plot axes) to be plotted by using the selected information. The optional flag following -B selects p(rimary) [Default] or s(econdary) axes information (mostly used for time axes annotations; see examples below). The components xinfo, yinfo and zinfo are of the form

info[:"axis label":][:="prefix":][:,"unit label":]

where info is one or more concatenated substrings of the form [t]stride[±phase][u]. The t flag sets the axis item of interest; the available items are listed in Table 4.2. By default, all 4 map boundaries (or plot axes) are plotted (denoted W, E, S, N). To change this selection, append the codes for those you want (e.g., WSn). Upper case (e.g., W) will annotate in addition to draw axis/tick-marks. The title, if given, will appear centered above the plot14. Unit label or prefix may start with a leading – to suppress the space between it and the annotation. Normally, equidistant annotations occur at multiples of stride; you can phase-shift this by appending ±phase.




FlagDescription


a Annotation tick spacing


f Frame tick spacing


g Grid tick spacing



Table 4.2: Interval type codes.

Note that the appearance of certain time annotations (month-, week-, and day-names) may be affected by the TIME_LANGUAGE, TIME_FORMAT_PRIMARY, and TIME_FORMAT_SECONDARY settings.

The unit flag u can take on one of 18 codes; these are listed in Table 4.3. Almost all of these units are time-axis specific. However, the m and c units will be interpreted as arc minutes and arc seconds, respectively, when a map projection is in effect.





FlagUnit Description



Y year Plot using all 4 digits



y year Plot using last 2 digits



O month Format annotation using PLOT_DATE_FORMAT



o month Plot as 2-digit integer (1–12)



U ISO week Format annotation using PLOT_DATE_FORMAT



u ISO week Plot as 2-digit integer (1–53)



r Gregorian week7-day stride from start of week (see TIME_WEEK_START)



K ISO weekday Plot name of weekday in selected language



k weekday Plot number of day in the week (1-7) (see TIME_WEEK_START)



D date Format annotation using PLOT_DATE_FORMAT



d day Plot day of month (1–31) or day of year (1–366)
(see PLOT_DATE_FORMAT



R day Same as d; annotations aligned with week (see TIME_WEEK_START)



H hour Format annotation using PLOT_CLOCK_FORMAT



h hour Plot as 2-digit integer (0–24)



M minute Format annotation using PLOT_CLOCK_FORMAT



m minute Plot as 2-digit integer (0–60)



C seconds Format annotation using PLOT_CLOCK_FORMAT



c seconds Plot as 2-digit integer (0–60)




Table 4.3: Interval unit codes.

There may be two levels of annotations. Here, “primary” refers to the annotation that is closest to the axis (this is the primary annotation), while “secondary” refers to the secondary annotation that is plotted further from the axis. The examples below will clarify what is meant. Note that the terms “primary” and “secondary” do not reflect any hierarchical order of units: The “primary” annotation interval is usually smaller (e.g., days) while the “secondary” annotation interval typically is larger (e.g., months).

Geographic basemaps

Geographic basemaps may differ from regular plot axis in that some projections support a “fancy” form of axis and is selected by the BASEMAP_TYPE setting. The annotations will be formatted according to the PLOT_DEGREE_FORMAT template and DEGREE_SYMBOL setting. A simple example of part of a basemap is shown in Figure 4.6.


PIC

Figure 4.6: Geographic map border using separate selections for annotation, frame, and grid intervals. Formatting of the annotation is controlled by the parameter PLOT_DEGREE_FORMAT in your .gmtdefaults4 file.

The machinery for primary and secondary annotations introduced for time-series axes can also be utilized for geographic basemaps. This may be used to separate degree annotations from minutes- and seconds-annotations. For a more complicated basemap example using several sets of intervals, including different intervals and pen attributes for grid lines and grid crosses, see Figure 4.7.


PIC

Figure 4.7: Geographic map border with both primary (P) and secondary (S) components.

Cartesian linear axes

For non-geographic axes, the BASEMAP_TYPE setting is implicitly set to plain. Other than that, cartesian linear axes are very similar to geographic axes. The annotation format may be controlled with the D_FORMAT parameter. By default, it is set to “%g”, which is a C language format statement for floating point numbers15, and with this setting the various axis routines will automatically determine how many decimal points should be used by inspecting the stride settings. If D_FORMAT is set to another format it will be used directly (.e.g, “%.2f” for a fixed, two decimals format). Note that for these axes you may use the unit setting to add a unit string to each annotation (see Figure 4.8).


PIC

Figure 4.8: Linear Cartesian projection axis. Long tickmarks accompany annotations, shorter ticks indicate frame interval. The axis label is optional. We used -R0/12/0/1 -JX3/0.4 -Ba4f2g1:Frequency::,%:.

Cartesian log10 axes

Due to the logarithmic nature of annotation spacings, the stride parameter takes on specific meanings. The following concerns are specific to log axes:

  1. stride must be 1, 2, 3, or a negative integer -n. Annotations/ticks will then occur at 1, 1–2–5, or 1,2,3,4,...,9, respectively, for each magnitude range. For -n the annotations will take place every n’th magnitude.
  2. Append l to stride. Then, log10 of the annotation is plotted at every integer log10 value (e.g., x = 100 will be annotated as “2”) [Default annotates x as is].
  3. Append p to stride. Then, annotations appear as 10 raised to log10 of the value (e.g., 10-5).

PIC

Figure 4.9: Logarithmic projection axis using separate values for annotation, frame, and grid intervals. (top) Here, we have chosen to annotate the actual values. Interval = 1 means every whole power of 10, 2 means 1, 2, 5 times powers of 10, and 3 means every 0.1 times powers of 10. We used -R1/1000/0/1 -JX3l/0.4 -Ba1f2g3. (middle) Here, we have chosen to annotate log10 of the actual values, with -Ba1f2g3l. (bottom) We annotate every power of 10 using log10 of the actual values as exponents, with -Ba1f2g3p.

Cartesian exponential axes

Normally, stride will be used to create equidistant (in the user’s unit) annotations or ticks, but because of the exponential nature of the axis, such annotations may converge on each other at one end of the axis. To avoid this problem, you can append p to stride, and the annotation interval is expected to be in transformed units, yet the annotation itself will be plotted as un-transformed units. E.g., if stride = 1 and power = 0.5 (i.e., sqrt), then equidistant annotations labeled 1, 4, 9, ... will appear.


PIC

Figure 4.10: Exponential or power projection axis. (top) Using an exponent of 0.5 yields a x axis. Here, intervals refer to actual data values, in -R0/100/0/1 -JX3p0.5/0.4 -Ba20f10g5. (bottom) Here, intervals refer to projected values, although the annotation uses the corresponding unprojected values, as in -Ba3f2g1p.

Cartesian time axes

What sets time axis apart from the other kinds of plot axes is the numerous ways in which we may want to tick and annotate the axis. Not only do we have both primary and secondary annotation items but we also have interval annotations versus tickmark annotations, numerous time units, and several ways in which to modify the plot. We will demonstrate this flexibility with a series of examples. While all our examples will only show a single x-axis, time-axis is supported for all axes.

Our first example shows a time period of almost two months in Spring 2000. We want to annotate the month intervals as well as the date at the start of each week:

_________________________________________________________________________________

gmtset PLOT_DATE_FORMAT -o ANNOT_FONT_SIZE_PRIMARY +9p  
psbasemap -R2000-4-1T/2000-5-25T/0/1 -JX5/0.2 -Bpa7Rf1d -Bsa1OS -P > GMT_-B_time1.ps
__________________________________________________________________________________________________

These commands result in Figure 4.11. Note the leading hyphen in the PLOT_DATE_FORMAT removes leading zeros from calendar items (e.g., 02 becomes 2).


PIC

Figure 4.11: Cartesian time axis, example 1.

The next example shows two different ways to annotate an axis portraying 2 days in July 1969:

_________________________________________________________________________________

gmtset PLOT_DATE_FORMAT ~o dd~ PLOT_CLOCK_FORMAT hh:mm ANNOT_FONT_SIZE_PRIMARY +9p  
psbasemap -R1969-7-21T/1969-7-23T/0/1 -JX5/0.2 -Bpa6Hf1h -Bsa1KS -P -K > GMT_-B_time2.ps  
psbasemap -R -J -Bpa6Hf1h -Bsa1DS -O -Y0.65i >> GMT_-B_time2.ps
__________________________________________________________________________________________________

The lower example (Figure 4.12) chooses to annotate the weekdays (by specifying a1K) while the upper example choses dates (by specifying a1D). Note how the clock format only selects hours and minutes (no seconds) and the date format selects a month name, followed by one space and a two-digit day-of-month number.


PIC

Figure 4.12: Cartesian time axis, example 2.

The third example presents two years, annotating both the years and every 3rd month.

_________________________________________________________________________________

gmtset PLOT_DATE_FORMAT o TIME_FORMAT_PRIMARY Character ANNOT_FONT_SIZE_PRIMARY +9p  
psbasemap -R1997T/1999T/0/1 -JX5/0.2 -Bpa3Of1o -Bsa1YS -P > GMT_-B_time3.ps
__________________________________________________________________________________________________

Note that while the year annotation is centered on the 1-year interval, the month annotations must be centered on the corresponding month and not the 3-month interval. The PLOT_DATE_FORMAT selects month name only and TIME_FORMAT_PRIMARY selects the 1-character, upper case abbreviation of month names using the current language (selected by TIME_LANGUAGE).


PIC

Figure 4.13: Cartesian time axis, example 3.

The fourth example (Figure 4.14) only shows a few hours of a day, using relative time by specifying t in the -R option while the TIME_UNIT is d (for days). We select both primary and secondary annotations, ask for a 12-hour clock, and let time go from right to left:

_________________________________________________________________________________

gmtset PLOT_CLOCK_FORMAT -hham ANNOT_FONT_SIZE_PRIMARY +9p  
psbasemap -R0.2t/0.35t/0/1 -JX-5/0.2 -Bpa15mf5m -Bsa1HS -P > GMT_-B_time4.ps
__________________________________________________________________________________________________

PIC

Figure 4.14: Cartesian time axis, example 4.

The fifth example shows a few weeks of time (Figure 4.15). The lower axis shows ISO weeks with week numbers and abbreviated names of the weekdays. The upper uses Gregorian weeks (which start at the day chosen by TIME_WEEK_START); they do not have numbers. ______________________________________________________________________________

gmtset PLOT_DATE_FORMAT u TIME_FORMAT_PRIMARY Character TIME_FORMAT_SECONDARY full \  
        ANNOT_FONT_SIZE_PRIMARY +9p  
psbasemap -R1969-7-21T/1969-8-9T/0/1 -JX5/0.2 -Bpa1K -Bsa1US -P -K > GMT_-B_time5.ps  
gmtset PLOT_DATE_FORMAT o TIME_WEEK_START Sunday TIME_FORMAT_SECONDARY Chararacter  
psbasemap -R -J -Bpa3Kf1k -Bsa1rS -O -Y0.65i >> GMT_-B_time5.ps
__________________________________________________________________________________________________

PIC

Figure 4.15: Cartesian time axis, example 5.

Our sixth example shows the first five months of 1996, and we have annotated each month with an abbreviated, upper case name and 2-digit year. Only the primary axes information is specified. ______

gmtset PLOT_DATE_FORMAT ~o yy~ TIME_FORMAT_PRIMARY Abbreviated  
psbasemap -R1996T/1996-6T/0/1 -JX5/0.2 -Ba1Of1dS -P > GMT_-B_time6.ps
__________________________________________________________________________________________________

PIC

Figure 4.16: Cartesian time axis, example 6.

Our seventh and final example illustrates annotation of year-days. Unless we specify the formatting with a leading hyphen in PLOT_DATE_FORMAT we get 3-digit integer days. Note that in order to have the two years annotated we need to allow for the annotation of small fractional intervals; normally such truncated interval must be at least half of a full interval. ______________________________________________________________________________________________________________________

gmtset PLOT_DATE_FORMAT jjj TIME_INTERVAL_FRACTION 0.05 ANNOT_FONT_SIZE_PRIMARY +9p  
psbasemap -R2000-12-15T/2001-1-15T/0/1 -JX5/0.2 -Bpa5Df1d -Bsa1YS -P > GMT_-B_time7.ps
__________________________________________________________________________________________________

PIC

Figure 4.17: Cartesian time axis, example 7.

4.4.4 Header data records: The -H option

The -H[i][n_recs] option lets GMT know that input file(s) have one [Default] or more header records. If there are more than one header record you must specify the number after the -H option, e.g., -H4. The default number of header records if -H is used is one of the many parameters in the .gmtdefaults4 file (N_HEADER_RECS), but can be overridden by -Hn_header_recs. Note that blank lines and records that be start with the character # are automatically skipped. Normally, programs that both read and write tables will output the header records that are found on input. Use -Hi to suppress the writing of header records.

4.4.5 Portrait plot orientation: The -P option


PIC

Figure 4.18: Users can specify Landscape [Default] or Portrait (-P) orientation.

-P selects Portrait plotting mode16. In general, a plot has an x-axis increasing from left to right and a y-axis increasing from bottom to top. If the paper is turned so that the long dimension of the paper is parallel to the x-axis then the plot is said to have Landscape orientation. If the long dimension of the paper parallels the y-axis the orientation is called Portrait (think of taking pictures with a camera and these words make sense). The default Landscape orientation is obtained by translating the origin in the x-direction (by the width of the chosen paper PAPER_MEDIA) and then rotating the coordinate system counterclockwise by 90°. By default the PAPER_MEDIA is set to Letter (or A4 if SI is chosen); this value must be changed when using different media, such as 11" x 17" or large format plotters (Figure 4.18).

4.4.6 Plot overlays: The -K-O options


PIC

Figure 4.19: A final PostScript file consists of any number of individual pieces.

The -K and -O options control the generation of PostScript code for multiple overlay plots. All PostScript files must have a header (for initializations), a body (drawing the figure), and a trailer (printing it out) (see Figure 4.19). Thus, when overlaying several GMT plots we must make sure that the first plot call omits the trailer, that all intermediate calls omit both header and trailer, and that the final overlay omits the header. -K omits the trailer which implies that more PostScript code will be appended later [Default terminates the plot system]. -O selects Overlay plot mode and omits the header information [Default initializes a new plot system]. Most unexpected results for multiple overlay plots can be traced to the incorrect use of these options. If you run only one plot program, ignore both the -O and -K options; they are only used when stacking plots.

4.4.7 Timestamps on plots: The -U option

-U draws UNIX System time stamp. Optionally, append an arbitrary text string (surrounded by double quotes), or the code c, which will plot the current command string (Figure 4.20).


PIC

Figure 4.20: The -U option makes it easy to “date” a plot.

4.4.8 Verbose feedback: The -V option

-V selects verbose mode, which will send progress reports to stderr [Default runs “silently”]. The interpretation of this option can be toggled by changing the default VERBOSE.

4.4.9 Plot positioning and layout: The -X-Y options


PIC

Figure 4.21: Plot origin can be translated freely with -X-Y.

-X and -Y shift origin of plot by (xoff ,yoff ) inches (Default is (X_ORIGIN, Y_ORIGIN) for new plots17 and (0,0) for overlays (-O)). By default, all translations are relative to the previous origin (see Figure 4.21). Supply offset as c to center the plot in that direction relative to the page margin. Absolute translations (i.e., relative to a fixed point (0,0) at the lower left corner of the paper) can be achieve by prepending “a” to the offsets. Subsequent overlays will be co-registered with the previous plot unless the origin is shifted using these options. The offsets are measured in the current coordinates system (which can be rotated using the initial -P option; subsequent -P options for overlays are ignored).

4.4.10 Binary table i/o: The -b option

All GMT programs that accept table data input may read ASCII, native binary, or netCDF data. When using native binary data the user must be aware of the fact that GMT has no way of determining the actual number of columns in the file. You must therefore pass that information to GMT via the binary -bi[s]n option, where n is the actual number of data columns (s indicates single (4 bytes) rather than double (8 bytes) precision). If uppercase S (or D) are used it implies that byte-swapping should be performed just prior to writing (for output) or immediately after reading (for input). Note that n may be larger than m, the number of columns that the GMT program requires to do its task. If n is not given then it defaults to m. If n < m an error is generated.

Because of its meta data, reading netCDF tables (i.e., netCDF files containing 1-dimensional arrays) is quite a bit less complex than reading native binary files. When feeding netCDF tables to programs like psxy, the program will automatically recognize the format and read whatever amount of columns are needed for that program. To steer which columns are to be read, the user can either append the suffix ?var1/var2/... to the netCDF file name or add the option -bicvar1/var2/..., where var1, var2, etc. are the names of the variables to be processed. The latter option is particularly practical when more than one file is read: the -bic option will apply to all files.

Currently, netCDF tables can only be input, not output. For more information, see Appendix B.

4.4.11 Number of Copies: The -c option

The -c option specifies the number of plot copies [Default is 1]. This value is embedded in the PostScript file and will make a printer issue the chosen number of copies without respooling.

4.4.12 Data type selection: The -f option

When map projections are not required we must explicitly state what kind of data each input or output column contains. This is accomplished with the -f option. Following an optional i (for input only) or o (for output only), we append a text string with information about each column (or range of columns) separated by commas. Each string starts with the column number (0 is first column) followed by either x (longitude), y (latitude), T (absolute calendar time) or t (relative time). If several consecutive columns have the same format you may specify a range of columns rather than a single column, i.e., 0-4 for the first 5 columns. For example, if our input file has geographic coordinates (latitude, longitude) with absolute calendar coordinates in the columns 3 and 4, we would specify fi0y,1x,3-4T. All other columns are assumed to have the default, floating point format and need not be set individually. The shorthand -f[i|o]g means -f[i|o]0x,1y (geographic coordinates). For more information, see Sections 4.10 and 4.11.

4.4.13 Data gap detection: The -g option

GMT has several mechanisms that can determine line segmentation. Typically, data segments are separated by multiple segment header records (see section 4.4.14 on -m below). However, if key data columns contain a NaN we may also use that information to break lines into multiple segments. This behavior is modified by the parameter NAN_RECORDS which by default is set to skip, meaning such records are considered bad and simply skipped. If you wish such records to indicate a segment boundary then set this parameter to pass. Finally, you may wish to indicate gaps based on the data values themselves. The -g option is used to detect gaps based on one or more criteria (use -g+ if all the criteria must be met; otherwise only one of the specified criteria needs to be met to signify a data gap). Gaps can be based on excessive jumps in the x- or y-coordinates (-gx or -gy), or on the distance between points (-gd). Append the gap distance and optionally a unit for actual distances. For geographic data the optional unit may be meter [Default], kilometer, miles, or nautical miles. For programs that maps data to map coordinates you can optionally specify these criteria to apply to the projected coordinates (by using upper-case -gX, -gY or-gD). In that case, choose from inch, centimeter, meter, or points. [Default unit is controlled by MEASURE_UNIT]. Note: For -gx or -gy with time data the unit is instead controlled by TIME_UNIT.

4.4.14 Multiple segment data: The -m option

The -m option states that the input and output table data will contain special records that marks the start of new line (or polygon) segments. These records are identified by their first character, which can be specified as an argument to -m [The default is >]. Append the modifiers i or o if the option should only apply to input or output, respectively. For binary data a multiple segment header is identified as a data record where all fields equal NaN. See section 4.4.14 and Appendix B for more details.

4.4.15 Lat/Lon or Lon/Lat?: The -: option

For geographical data, the first column is expected to contain longitudes and the second to contain latitudes. To reverse this expectation you must apply the -: option. Optionally, append i or o to restrict the effect to input or output only. Note that command line arguments that may take geographic coordinates (e.g., -R) always expect longitude before latitude. Also, geographical grids are expected to have the longitude as first (minor) dimension.

4.5 Command line history

GMT programs “remember” the standardized command line options (See Section 4.4) given during their previous invocations and this provides a shorthand notation for complex options. For example, if a basemap was created with an oblique Mercator projection, specified as

-Joc170W/25:30S/33W/56:20N/1:500000

then a subsequent psxy command to plot symbols only needs to state -Jo in order to activate the same projection. In contrast, note that -J by itself will pick the most recently used projection. Previous commands are maintained in the file .gmtcommands4, of which there will be one in each directory you run the programs from. This is handy if you create separate directories for separate projects since chances are that data manipulations and plotting for each project will share many of the same options. Note that an option spelled out on the command line will always override the last entry in the .gmtcommands4 file and, if execution is successful, will replace this entry as the previous option argument in the .gmtcommands4 file. If you call several GMT modules piped together then GMT cannot guarantee that the .gmtcommands4 file is processed in the intended order from left to right. The only guarantee is that the file will not be clobbered since GMT uses advisory file locking. The uncertainty in processing order makes the use of shorthands in pipes unreliable. We therefore recommend that you only use shorthands in single process command lines, and spell out the full command option when using chains of commands connected with pipes.

4.6 Usage messages, syntax- and general error messages

Each program carries a usage message. If you enter the program name without any arguments, the program will write the complete usage message to standard error (your screen, unless you redirect it). This message explains in detail what all the valid arguments are. If you enter the program name followed by a hyphen (–) only you will get a shorter version which only shows the command line syntax and no detailed explanations. If you incorrectly specify an option or omit a required option, the program will produce syntax errors and explain what the correct syntax for these options should be. If an error occurs during the running of a program, the program will in some cases recognize this and give you an error message. Usually this will also terminate the run. The error messages generally begin with the name of the program in which the error occurred; if you have several programs piped together this tells you where the trouble is.

4.7 Standard input or file, header records

Most of the programs which expect table data input can read either standard input or input in one or several files. These programs will try to read stdin unless you type the filename(s) on the command line without the above hyphens. (If the program sees a hyphen, it reads the next character as an instruction; if an argument begins without a hyphen, it tries to open this argument as a filename). This feature allows you to connect programs with pipes if you like. If your input is ASCII and has one or more header records, you must use the -H option (see Section 4.4.4). For binary table data no headers are allowed. ASCII files may in many cases also contain sub-headers separating data segments. These are called “multi-segment files” and requires a special option (typically -m); see Appendix B for complete documentation.

If filenames are given for reading, GMT programs will first look for them in the current directory. If the file is not found, the programs will look in two other directories pointed to by environmental parameters (if set). These are GMT_DATADIR and GMT_USERDIR, and they may be set by the user to point to directories that contain data sets of general use. Normally, the first directory (or directories: add multiple paths by separating them with colons (semi-colons under Windows)) will hold data sets of a general nature (tables, grids), although a particular use is to make available large grids accessible via the supplemental programs grdraster or img2grd; see Appendix A for information about these supplemental programs. The second directory may hold miscellaneous data sets more specific to the user; this directory also stores GMT defaults and other configuration files. Data sets that the user finds are often needed may be placed in these directories, thus eliminating the need to specify a full path to the file. Program output is always written to the current directory unless a full path has been specified.

4.8 Verbose operation

Most of the programs take an optional -V argument which will run the program in the “verbose” mode (see Section 4.4.8). Verbose will write to standard error information about the progress of the operation you are running. Verbose reports things such as counts of points read, names of data files processed, convergence of iterative solutions, and the like. Since these messages are written to stderr, the verbose talk remains separate from your data output.

4.9 Program output

Most programs write their results, including PostScript plots, to standard output. The exceptions are those which may create binary netCDF grid files such as surface (due to the design of netCDF a filename must be provided; however, alternative binary output formats allowing piping are available; see Section 4.17). With UNIX you can redirect standard output to a file or pipe it into another process. Error messages, usage messages, and verbose comments are written to standard error in all cases. You can use UNIX to redirect standard error as well, if you want to create a log file of what you are doing.

4.10 Input data formats

Most of the time, GMT will know what kind of x and y coordinates it is reading because you have selected a particular coordinate transformation or map projection. However, there may be times when you must explicitly specify what you are providing as input using the -f switch. When binary input data are expected (-bi) they must all be floating point numbers, however for ASCII input there are numerous ways to encode data coordinates (which may be separated by white-space or commas). Valid input data are generally of the same form as the arguments to the -R option (see Section 4.4.1), with additional flexibility for calendar data. Geographical coordinates, for example, can be given in decimal degrees (e.g., -123.45417) or in the [±]ddd[:mm[:ss[.xxx]]][W|E|S|N] format (e.g., 123:27:15W).

Because of the widespread use of incompatible and ambiguous formats, the processing of input date components is guided by the template INPUT_DATE_FORMAT in your .gmtdefaults4 file; it is by default set to yyyy-mm-dd. Y2K-challenged input data such as 29/05/89 can be processed by setting INPUT_DATE_FORMAT to dd/mm/yy. A complete description of possible formats is given in the gmtdefaults man page. The clock string is more standardized but issues like 12- or 24-hour clocks complicate matters as well as the presence or absence of delimiters between fields. Thus, the processing of input clock coordinates is guided by the template INPUT_CLOCK_FORMAT which defaults to hh:mm:ss.xxx.

GMT programs that require a map projection argument will implicitly know what kind of data to expect, and the input processing is done accordingly. However, some programs that simply report on minimum and maximum values or just do a reformatting of the data will in general not know what to expect, and furthermore there is no way for the programs to know what kind of data other columns (beyond the leading x and y columns) contain. In such instances we must explicitly tell GMT that we are feeding it data in the specific geographic or calendar formats (floating point data are assumed by default). We specify the data type via the -f option (which sets both input and output formats; use -fi and -fo to set input and output separately). For instance, to specify that the the first two columns are longitude and latitude, and that the third column (e.g., z) is absolute calendar time, we add -fi0x,1y,2T to the command line. For more details, see the man page for the program you need to use.

4.11 Output data formats

The numerical output from GMT programs can be binary (when -bo is used) or ASCII [Default]. In the latter case the issue of formatting becomes important. GMT provides extensive machinery for allowing just about any imaginable format to be used on output. Analogous to the processing of input data, several templates guide the formatting process. These are OUTPUT_DATE_FORMAT and OUTPUT_CLOCK_FORMAT for calendar-time coordinates, OUTPUT_DEGREE_FORMAT for geographical coordinates, and D_FORMAT for generic floating point data. In addition, the user have control over how columns are separated via the FIELD_SEPARATOR parameter. Thus, as an example, it is possible to create limited FORTRAN-style card records by setting D_FORMAT to %7.3lf and FIELD_SEPARATOR to none [Default is tab].

4.12 PostScript features

PostScript is a command language for driving graphics devices such as laser printers. It is ASCII text which you can read and edit as you wish (assuming you have some knowledge of the syntax). We prefer this to binary metafile plot systems since such files cannot easily be modified after they have been created. GMT programs also write many comments to the plot file which make it easier for users to orient themselves should they need to edit the file (e.g., % Start of x-axis). All GMT programs create PostScript code by calling the pslib plot library (The user may call these functions from his/her own C or FORTRAN plot programs. See the manual pages for pslib syntax). Although GMT programs can create very individualized plot code, there will always be cases not covered by these programs. Some knowledge of PostScript will enable the user to add such features directly into the plot file. By default, GMT will produce freeform PostScript output with embedded printer directives. To produce Encapsulated PostScript (EPS) that can be imported into graphics programs such as IslandDraw, CorelDraw, Illustrator or Freehand for further embellishment, change the PAPER_MEDIA setting in the .gmtdefaults4 file. See Appendix C and the gmtdefaults man page for more details.

4.13 Specifying pen attributes

A pen in GMT has three attributes: width, color, and texture. Most programs will accept pen attributes in the form of an option argument, with commas separating the given attributes, e.g.,

-W[width[c|i|p|m]],[color],[texture[c|i|p|m]]

Width is by default measured in units of the current device resolution (i.e., the value assigned to the parameter DOTS_PR_INCH in your .gmtdefaults4 file). Thus, if the dpi is set to 300 this unit is 1/300th of an inch. Append c, i, p, or m to specify pen width in cm, inch, points (1/72 of an inch), or meters, respectively. Note that a pen thickness of 5 will be of different physical width depending on your dpi setting, whereas a thickness of 5p will always be 5/72 of an inch. Minimum-thickness pens can be achieved by giving zero width, but the result is device-dependent. Finally, a few predefined pen names can be used: default, faint, and {thin, thick, fat}[er|est], and obese. Table 4.4 shows this list and the corresponding pen widths.






Pen name
Width
Pen name
Width




faint 0 thicker 1.5p




default 0.25pthickest 2p




thinnest 0.25pfat 3p




thinner 0.50pfatter 6p




thin 0.75pfattest 12p




thick 1.0p obese 18p





Table 4.4: GMT predefined pen widths.

The color can be specified in five different ways:
  1. Gray. Specify a gray shade in the range 0–255 (linearly going from black [0] to white [255]).
  2. RGB. Specify r/g/b, each ranging from 0–255. Here 0/0/0 is black, 255/255/255 is white, 255/0/0 is red, etc.
  3. HSV. Specify hue-saturation-value, with the former in the 0–360 degree range while the latter two take on the range 0–118.
  4. CMYK. Specify cyan/magenta/yellow/black, each ranging from 0–100%.
  5. Name. Specify one of 663 valid color names. Use man gmtcolors to list all valid names. A very small yet versatile subset consists of the 29 choices white, black, and [light|dark]{red, orange, yellow, green, cyan, blue, magenta, gray|grey, brown}. The color names are case-insensitive, so mixed upper and lower case can be used (like DarkGreen).
The texture attribute controls the appearance of the line. A “.” yields a dotted line, whereas a dashed pen is requested with “-”. Also combinations of dots and dashes, like “.-” for a dot-dashed line, are allowed. The lengths of dots and dashes are scaled relative to the pen width (dots has a length that equals the pen width while dashes are 8 times as long; gaps between segments are 4 times the pen width). For more detailed attributes including exact dimensions you may specify string:offset, where string is a series of numbers separated by underscores. These numbers represent a pattern by indicating the length of line segments and the gap between segments. The offset phase-shifts the pattern from the beginning the line. For example, if you want a yellow line of width 0.1 cm that alternates between long dashes (4 points), an 8 point gap, then a 5 point dash, then another 8 point gap, with pattern offset by 2 points from the origin, specify -W0.1c,yellow,4_8_5_8:2p. In general, the texture units can be specified in dpi units, cm, inch, points, or meters (see width discussion above).

Table 4.5 contains additional examples of pen specifications suitable for, say, psxy.




Pen example
Comment


-W0.5p Solid black line, 0.5 point thick


-Wgreen Solid green line with default width


-Wthin,red,- Dashed, thin red line


-Wfat,. Fat dotted line [black]


-W0.1c,120-1-1 Green (in h-s-v) pen, 1 mm thick


-Wfaint,100/0/0/0,..-Very thin, cyan (in c/m/y/k), dot-dot-dashed line



Table 4.5: A few examples of pen specifications.

4.14 Specifying area fill attributes

Many plotting programs will allow the user to draw filled polygons or symbols. The fill specification may take two forms:

-Gfill

-Gpdpi/pattern[:Bcolor[Fcolor]]

fill:
In the first case we may specify a gray shade (0–255), RGB color (r/g/b all in the 0–255 range or in hexadecimal #rrggbb), HSV color (hue-saturation-value in the 0–360, 0–1, 0–1 range), CMYK color (cyan/magenta/yellow/black, each ranging from 0–100%), or a valid color name; in that respect it is similar to specifying the pen color settings (see pen color discussion under Section 4.13).
pattern:
The second form allows us to use a predefined bit-image pattern. pattern can either be a number in the range 1–90 or the name of a 1-, 8-, or 24-bit Sun raster file. The former will result in one of the 90 predefined 64 x 64 bit-patterns provided with GMT and reproduced in Appendix E. The latter allows the user to create customized, repeating images using standard Sun raster files19. The dpi parameter sets the resolution of this image on the page; the area fill is thus made up of a series of these “tiles”. Specifying dpi as 0 will result in highest resolution obtainable given the present dpi setting in .gmtdefaults4. By specifying upper case -GP instead of -Gp the image will be bit-reversed, i.e., white and black areas will be interchanged (only applies to 1-bit images or predefined bit-image patterns). For these patterns and other 1-bit images one may specify alternative background and foreground colors (by appending :Bcolor[Fcolor]) that will replace the default white and black pixels, respectively. Setting one of the fore- or background colors to – yields a transparent image where only the back- or foreground pixels will be painted.

Due to PostScript implementation limitations the raster images used with -G must be less than 146 x 146 pixels in size; for larger images see psimage. The format of Sun raster files is outlined in Appendix B. Note that under PostScript Level 1 the patterns are filled by using the polygon as a clip path. Complex clip paths may require more memory than the PostScript interpreter has been assigned. There is therefore the possibility that some PostScript interpreters (especially those supplied with older laserwriters) will run out of memory and abort. Should that occur we recommend that you use a regular grayshade fill instead of the patterns. Installing more memory in your printer may or may not solve the problem!

Table 4.6 contains a few examples of fill specifications.




Fill example
Comment


-G128 Solid gray


-G127/255/0 Chartreuse, R/G/B-style


-G#00ff00 Green, hexadecimal RGB code


-G25-0.86-0.82 Chocolate, h-s-v – style


-GDarkOliveGreen1One of the named colors


-Gp300/7 Simple diagonal hachure pattern in b/w at 300 dpi


-Gp300/7:Bred Same, but with red lines on white


-Gp300/7:BredF- Now the gaps between red lines are transparent


-Gp100/marble.ras Using user image of marble as the fill at 100 dpi



Table 4.6: A few examples of fill specifications.

4.15 Color palette tables

Several programs, such as those which read 2-D gridded data sets and create colored images or shaded reliefs, need to be told what colors to use and over what z-range each color applies. This is the purpose of the color palette table (cpt-file). These files may also be used by psxy and psxyz to plot color-filled symbols. For most applications, you will simply create a cpt-file using the tool makecpt which will take an existing color table and resample it to fit your chosen data range, or use grd2cpt to build a cpt-file based on the data distribution in one or more given grid files. However, in some situations you will need to make a cpt-file by hand or using text tools like awk or perl.

Color palette tables (CPT) comes in two flavors: (1) Those designed to work with categorical data (e.g., data where interpolation of values is undefined) and (2) those designed for regular, continuously-varying data.

4.15.1 Categorical CPT files

Note: This is an experimental component and is only available if you compile GMT with -DGMT_CPT2. Categorical data are information on which normal numerical operations are not defined. As an example, consider various land classifications (desert, forest, glacier, etc.) and it is clear that even if we assigned a numerical value to these categories (e.g., desert = 1, forest = 2, etc) it would be meaningless to compute average values (what would 1.5 mean?). For such data a special format of the CPT files are provided. Here, each category is assigned a unique key, a color or pattern, and an optional label (usually the category name). Keys must be monotonically increasing but do not need to be consecutive. The format is

key1filllabel
keyn filllabel

The fill information follows the format given in Section 4.14. While not always applicable to categorical data, the background color (for key-values < key1), foreground color (for key-values > keyn), and not-a-number (NaN) color (for key-values = NaN) are all defined in the .gmtdefaults4 file, but can be overridden by the statements

BRbackGbackBback
F RforeGforeBfore
N RnanGnanBnan

4.15.2 Regular CPT files

Here, the colors may be specified either in the RGB- (red, green, blue), CMYK- (cyan, magenta, yellow, black), or in the HSV-system (hue, saturation, value, and here the comment # COLOR_MODEL = HSV must be present in the cpt file since there are no other way to distinguish between HSV and RGB). Color names can also be used. Using the RGB system20, the format of the cpt-file is:

z0RminGminBminz1RmaxGmaxBmax[A][;label]
zn-2 RminGminBminzn-1RmaxGmaxBmax[A][;label]

Thus, for each “z-slice”, defined as the interval between two boundaries (e.g., z0 to z1), the color can be constant (by letting Rmin = Rmax, Gmin = Gmax, and Bmin = Bmax) or a continuous, linear function of z. The optional flag A is used to indicate annotation of the color scale when plotted using psscale. The optional code A may be L, U, or B to select annotation of the lower, upper, or both limits of the particular z-slice. However, the standard -B option can be used by psscale to affect annotation and ticking of color scales. The optional semicolon followed by a text label will make psscale, when used with the -L option, place the supplied label instead of formatted z-values.

As for categorical tables, the background color (for z-values < z0), foreground color (for z-values > zn-1), and not-a-number (NaN) color (for z-values = NaN) are all defined in the .gmtdefaults4 file, but can be overridden by the statements

BRbackGbackBback
F RforeGforeBfore
N RnanGnanBnan

which can be inserted into the beginning or end of the cpt-file. If you prefer the HSV system, set the .gmtdefaults4 parameter accordingly and replace red, green, blue with hue, saturation, value. Color palette tables that contain grayshades only may replace the r/g/b triplets with a single grayshade in the 0–255 range. For CMYK, give four values in the 0–100 range. Both the min and max color specifications in one z-slice must use the same color system, i.e., you cannot mix “red” and 0/255/100 on the same line.

A few programs (i.e., those that plot polygons such as grdview, psscale, and psxy) can accept pattern fills instead of grayshades. You must specify the pattern as in Section 4.14 (no leading -G of course), and only the first (low z) is used (we cannot interpolate between patterns). Finally, some programs let you skip features whose z-slice in the cptfile has grayshades set to –. As an example, consider

30 p200/1680
80 100
100200 0 0 2002552550
200yellow 300green

where slice 30 < z < 80 is painted with pattern # 16 at 200 dpi, slice 80 < z < 100 is skipped, slice 100 < z < 200 is painted in a range of dark red to yellow, whereas the slice 200 < z < 300 will linearly yield colors from yellow to green, depending on the actual value of z.

Some programs like grdimage and grdview apply artificial illumination to achieve shaded relief maps. This is typically done by finding the directional gradient in the direction of the artificial light source and scaling the gradients to have approximately a normal distribution on the interval [-1,+1]. These intensities are used to add “white” or “black” to the color as defined by the z-values and the cpt-file. An intensity of zero leaves the color unchanged. Higher values will brighten the color, lower values will darken it, all without changing the original hue of the color (see Appendix I for more details). The illumination is decoupled from the data grid file in that a separate grid file holding intensities in the [-1,+1] range must be provided. Such intensity files can be derived from the data grid using grdgradient and modified with grdhisteq, but could equally well be a separate data set. E.g., some side-scan sonar systems collect both bathymetry and backscatter intensities, and one may want to use the latter information to specify the illumination of the colors defined by the former. Similarly, one could portray magnetic anomalies superimposed on topography by using the former for colors and the latter for shading.

4.16 Character escape sequences

For annotation labels or text strings plotted with pstext, GMT provides several escape sequences that allow the user to temporarily switch to the symbol font, turn on sub- or superscript, etc., within words. These conditions are toggled on/off by the escape sequence @x, where x can be one of several types. The escape sequences recognized in GMT are listed in Table 4.7. Only one level of sub- or superscript is supported. Note that under Windows the percent symbol indicates a batch variable, hence you must use two percent-signs for each one required in the escape sequence for font switching.




Code
Effect


Turns symbol font on or off


@+ Turns superscript on or off


@- Turns subscript on or off


@# Turns small caps on or off


@_ Turns underline on or off


@%fontno%Switches to another font; @%% resets to previous font


@:size: Switches to another font size; @:: resets to previous size


@;color; Switches to another font color; @;; resets to previous color


@! Creates one composite character of the next two characters


@@ Prints the @ sign itself



Table 4.7: GMT text escape sequences.

Shorthand notation for a few special European characters has also been added (Table 4.8):






CodeEffectCodeEffect




@E Æ @e æ




@O Ø @o ø




@A Å @a å




@C Ç @c ç




@N Ñ @n ñ




@U Ü @u ü




@s ß





Table 4.8: Shortcuts for some European characters.

PostScript fonts used in GMT may be re-encoded to include several accented characters used in many European languages. To access these, you must specify the full octal code \xxx allowed for your choice of character encodings determined by the CHAR_ENCODING setting described in the gmtdefaults man page. Only the special characters belonging to a particular encoding will be available. Many characters not directly available by using single octal codes may be constructed with the composite character mechanism @!.

Some examples of escape sequences and embedded octal codes in GMT strings using the Standard+ encoding:

2@~p@~r@+2@+h@-0@- E\363tv\363s = 2πr2h0 Eötvös

10@+-3 @Angstr@om = 10-3 Ångstrøm

Se@nor Gar@con = Señor Garçon

M@!\305anoa stra@se = Manoa straße

A@\#cceleration@\# (ms@+-2@+) = ACCELERATION (MS-2)

The option in pstext to draw a rectangle surrounding the text will not work for strings with escape sequences. A chart of characters and their octal codes is given in Appendix F.

4.17 Grid file format specifications

GMT has the ability to read and write grids using more than one grid file format (see Table 4.9 for supported format and their IDs). For reading, GMT will automatically determine the format of grid files, while for writing you will normally have to append =ID to the filename if you want GMT to use a different format than the default.

By default, GMT will create new grid files using the nf format; however, this behavior can be overridden by setting the GRID_FORMAT defaults parameter to any of the other recognized values (or by appending =ID).

GMT can also read netCDF grid files produced by other software packages, provided the grid files satisfy the COARDS and Hadley Centre conventions for netCDF grids. Thus, products created under those conventions (provided the grid is 2-, 3-, 4-, or 5-dimensional) can be read directly by GMT and the netCDF grids written by GMT can be read by other programs that conform to those conventions. Three such programs are ncview, Panoply and ncBrowse; others can be found on the netCDF website.

In addition, users with some C-programming experience may add their own read/write functions and link them with the GMT library to extend the number of predefined formats. Technical information on this topic can be found in the source file gmt_customio.c. Users who are considering this approach should contact the GMT team.




ID
GMT 4 netCDF standard formats




nb GMT netCDF format (byte) (COARDS-compliant)


ns GMT netCDF format (short) (COARDS-compliant)


ni GMT netCDF format (int) (COARDS-compliant)


nf GMT netCDF format (float) (COARDS-compliant)


nd GMT netCDF format (double) (COARDS-compliant)




ID
GMT 3 netCDF legacy formats




cb GMT netCDF format (byte) (depreciated)


cs GMT netCDF format (short) (depreciated)


ci GMT netCDF format (int) (depreciated)


cf GMT netCDF format (float) (depreciated)


cd GMT netCDF format (double) (depreciated)




ID
GMT native binary formats




bmGMT native, C-binary format (bit-mask)


bb GMT native, C-binary format (byte)


bs GMT native, C-binary format (short)


bi GMT native, C-binary format (int)


bf GMT native, C-binary format (float)


bd GMT native, C-binary format (double)




ID
Miscellaneous grid formats




rb SUN raster file format (8-bit standard)


rf GEODAS grid format GRD98 (NGDC)


sf Golden Software Surfer format 6 (float)


sd Golden Software Surfer format 7 (double)


af Atlantic Geoscience Center AGC (float)


gd Read-only via GDAL [EXPERIMENTAL] (float)



Table 4.9: GMT grid file formats.

Because some formats have limitations on the range of values they can store it is sometimes necessary to provide more than simply the name of the file and its ID on the command line. For instance, a native short integer file may use a unique value to signify an empty node or NaN, and the data may need translation and scaling prior to use. Therefore, all GMT programs that read or write grid files will decode the given filename as follows:

name[=ID[/scale/offset[/nan]]]

where everything in brackets is optional. If you are reading a grid then no options are needed: just continue to pass the name of the grid file. However, if you write another format you must append the =ID string, where ID is the format code listed above. In addition, should you want to (1) multiply the data by a scale factor, and (2) add a constant offset you must append the /scale/offset modifier. Finally, if you need to indicate that a certain data value should be interpreted as a NaN (not-a-number) you must append the /nan suffix to the scaling string (it cannot go by itself; note the nesting of the brackets!).

Some of the grid formats allow writing to standard output and reading from standard input which means you can connect GMT programs that operate on grid files with pipes, thereby speeding up execution and eliminating the need for large, intermediate grid files. You specify standard input/output by leaving out the filename entirely. That means the “filename” will begin with “=ID ” since no GMT netCDF format allow piping (due to the design of netCDF).

Everything looks clearer after a few examples:

  1. To write a native binary float grid file, specify the name as my_file.f4=bf.
  2. To read a native short integer grid file, multiply the data by 10 and then add 32000, but first let values that equal 32767 be set to NaN, use the filename my_file.i2=bs/10/32000/32767.
  3. To read a Golden Software “surfer” format 6 grid file, just pass the file name, e.g., my_surferfile.grd.
  4. To read a 8-bit standard Sun raster file (with values in the 0–255 range) and convert it to a ±1 range, give the name as rasterfile=rb/7.84313725e-3/-1 (i.e., 1/127.5).
  5. To write a native binary short integer grid file to standard output after subtracting 32000 and dividing its values by 10, give filename as =bs/0.1/-3200.

Programs that both read and/or write more than one grid file may specify different formats and/or scaling for the files involved. The only restriction with the embedded grid specification mechanism is that no grid files may actually use the “=” character as part of their name (presumably, a small sacrifice).

One can also define special file suffixes to imply a specific file format; this approach represents a more intuitive and user-friendly way to specify the various file formats. The user may create a file called .gmt_io in the current directory, home directory or in the directory /̃.gmt and define any number of custom formats. The following is an example of a .gmt_io file:

# GMT i/o shorthand file

# It can have any number of comment lines like this one anywhere

# suffix format_id scale offset NaN Comments

grd nf - - - Default format

b bf - - - Native binary floats

i2 bs - - 32767 2-byte integers with NaN value

ras rb - - - Sun raster files

byte bb - - 255 Native binary 1-byte grids

bit bm - - - Native binary 0 or 1 grids

mask bm - - 0 Native binary 1 or NaN masks

faa bs 0.1 - 32767 Native binary gravity in 0.1 mGal

These suffices can be anything that makes sense to the user. To activate this mechanism, set parameter GRIDFILE_SHORTHAND to TRUE in your .gmtdefaults4 file. Then, using the filename stuff.i2 is equivalent to saying stuff.i2=bs/1/0/32767, and the filename wet.mask means wet.mask=bm/1/0/0. For a file intended for masking, i.e., the nodes are either 1 or NaN, the bit or mask format file may be as small as 1/32 the size of the corresponding grid float format file.

4.18 Options for COARDS-compliant netCDF files

When the netCDF file contains more than one 2-dimensional variable, GMT programs will load the first such variable in the file and ignore all others. Alternatively, the user can select the required variable by adding the suffix “?varname” to the file name. For example, to get information on the variable “slp” in file file.nc, use:

grdinfo ~file.nc?slp~

Since COARDS-compliant netCDF files are the default, the additional suffix “=nf” can be omitted.

In case the named variable is 3-dimensional, GMT will load the first (bottom) layer. If another layer is required, either add “[index]” or “(level)”, where index is the index of the third (depth) variable (starting at 0 for the first layer) and level is the numerical value of the third (depth) variable associated with the requested layer. To indicate the second layer of the 3-D variable “slp” use as file name: file.nc?slp[1].

When you supply the numerical value for the third variable using “(level)”, GMT will pick the layer closest to that value. No interpolation is performed.

Note that the question mark, brackets and parentheses have special meanings on Unix-based platforms. Therefore, you will need to either escape these characters, by placing a backslash in front of them, or place the whole file name plus modifiers between single quotes or double quotes.

A similar approach is followed for loading 4-dimensional grids. Consider a 4-dimensional grid with the following variables:

lat(lat): 0, 1, 2, 3, 4, 5, 6, 7, 8, 9  
lon(lon): 0, 1, 2, 3, 4, 5, 6, 7, 8, 9  
depth(depth): 0, 10, 20, 30, 40, 50, 60, 70, 80, 90  
time(time): 0, 12, 24, 36, 48  
pressure(time,depth,lat,lon): (5000 values)

To get information on the 10×10 grid of pressure at depth 10 and at time 24, one would use:

grdinfo ~file.nc?pressure[2,1]~

or (only in case the coordinates increase linearly):

grdinfo ~file.nc?pressure(24,10)~

The COARDS conventions set restrictions on the names that can be used for the units of the variables and coordinates. For example, the units of longitude and latitude are “degrees_east” and “degrees_north”, respectively. Here is an example of the header of a COARDS compliant netCDF file (to be obtained using ncdump):

netcdf M2_fes2004 {  
dimensions:  
        lon = 2881 ;  
        lat = 1441 ;  
variables:  
        float lon(lon) ;  
                lon:long_name = ~longitude~ ;  
                lon:units = ~degrees_east~ ;  
                lon:actual_range = 0., 360. ;  
        float lat(lat) ;  
                lat:long_name = ~latitude~ ;  
                lat:units = ~degrees_north~ ;  
                lat:actual_range = -90., 90. ;  
        short amp(lat, lon) ;  
                amp:long_name = ~amplitude~ ;  
                amp:unit = ~m~ ;  
                amp:scale_factor = 0.0001 ;  
                amp:add_offset = 3. ;  
                amp:_FillValue = -32768s ;  
        short pha(lat, lon) ;  
                pha:long_name = ~phase~ ;  
                pha:unit = ~degrees~ ;  
                pha:scale_factor = 0.01 ;  
                pha:_FillValue = -32768s ;

This file contains two grids, which can be plotted separately using the names M2_fes2004.nc?amp and M2_fes2004.nc?pha. The attributes long_name and unit for each variable are combined in GMT to a single unit string. For example, after reading the grid y_unit equals latitude [degrees_north]. The same method can be used in reverse to set the proper variable names and units when writing a grid. However, when the coordinates are set properly as geographical or time axes, GMT will take care of this. The user is, however, still responsible for setting the variable name and unit of the z-coordinate. The default is simply “z”.

4.19 The NaN data value

For a variety of data processing and plotting tasks there is a need to acknowledge that a data point is missing or unassigned. In the “old days” such information was passed by letting a value like -9999.99 take on the special meaning of “this is not really a value, it is missing”. The problem with this scheme is that -9999.99 (or any other floating point value) may be a perfectly reasonable data value and in such a scenario would be skipped. The solution adopted in GMT is to use the IEEE concept Not-a-Number (NaN) for this purpose. Mathematically, a NaN is what you get if you do an undefined mathematical operation like 00; in ASCII data files they appear as the textstring NaN. This value is internally stored with a particular bit pattern defined by IEEE so that special action can be taken when it is encountered by programs. In particular, a library function called isnan is used to test if a floating point is a NaN. GMT uses these tests extensively to determine if a value is suitable for plotting or processing (if a NaN is used in a calculation the result would become NaN as well). Data points whose values equal NaN are not normally plotted (or plotted with the special NaN color given in .gmtdefaults4). Several tools such as xyz2grd, gmtmath, and grdmath can convert user data to NaN and vice versa, thus facilitating arbitrary masking and clipping of data sets. Note that a few computers do not have native IEEE hardware support. At this point, this applies to some of the older Cray super-computers. Users on such machines may have to adopt the old ‘-9999.99” scheme to achieve the desired results.

Data records that contain NaN values for the x or y columns (or the z column for cases when 3-D Cartesian data are expected) are usually skipped during reading. However, the presence of these bad records can be interpreted in two different ways, and this behavior is controlled by the NAN_RECORDS defaults parameter. The default setting (gap) considers such records to indicate a gap in an otherwise continuous series of points (e.g., a line), and programs can act upon this information, e.g., not to draw a line across the gap or to break the line into separate segments. The alternative setting (bad) makes no such interpretation and simply reports back how many bad records were skipped during reading.

4.20 GMT environment parameters

GMT relies on several environment parameters, in particular to find data files and program settings.

$GMT_SHAREDIR
points to the GMT share directory where all run-time support files such as coastlines, custom symbols, PostScript macros, color tables, and much more reside. If this parameter is not set it defaults to the share sub-directory selected during the GMT install process (e.g., your answer to question C.9 on the web install form). If no selection was made the ultimate default is the share directory under the GMT installation directory.
$GMT_DATADIR
points to one or more directory where large and/or widely used data files can be placed. All GMT programs look in this directories when a file is specified on the command line and it is not present in the current directory. This allows maintainers to consolidate large data files and to simplify scripting that use these files since the absolute path need not be specified. Separate multiple directories with colons; under Windows you use semi-colons.
$GMT_USERDIR
points to a directory where the user may place custom configuration files (e.g., an alternate coastline.conf file, preferred default settings in .gmtdefaults4, custom symbols and color palettes, and shorthands for gridfile extensions via .gmt_io). Users may also place their own data files in this directory as GMT programs will search for files given on the command line in both $GMT_DATADIR and $GMT_USERDIR.
$GMT_TMPDIR
is where GMT will write its state parameters via the three files .gmtcommands4, .gmtdefaults4 and .gmt_bb_info. If not set then these files are written to the current directory. See Appendix P for more on the use of $GMT_TMPDIR.

Chapter 5
GMT Coordinate Transformations

GMT programs read real-world coordinates and convert them to positions on a plot. This is achieved by selecting one of several coordinate transformations or projections. We distinguish between three sets of such conversions:

The next chapter will be dedicated to GMT map projections in its entirety. Meanwhile, the present chapter will summarize the properties of the Cartesian and Polar coordinate transformations available in GMT, list which parameters define them, and demonstrate how they are used to create simple plot axes. We will mostly be using psbasemap (and occasionally psxy) to demonstrate the various transformations. Our illustrations may differ from those you reproduce with the same commands because of different settings in our .gmtdefaults4 file.) Finally, note that while we will specify dimensions in inches (by appending i), you may want to use cm (c), meters (m), or points (p) as unit instead (see the gmtdefaults man page).

5.1 Cartesian transformations

GMT Cartesian coordinate transformations come in three flavors:

These transformations convert input coordinates (x,y) to locations (x,y) on a plot. There is no coupling between x and y (i.e., x = f(x) and y = f(y)); it is a one-dimensional projection. Hence, we may use separate transformations for the x- and y-axes (and z-axes for 3-D plots). Below, we will use the expression u = f(u), where u is either x or y (or z for 3-D plots). The coefficients in f(u) depend on the desired plot size (or scale), the chosen (x,y) domain, and the nature of f itself.

Two subsets of linear will be discussed separately; these are a polar (cylindrical) projection and a linear projection applied to geographic coordinates (with a 360° periodicity in the x-coordinate). We will show examples of all of these projections using dummy data sets created with gmtmath, a “Reverse Polish Notation” (RPN) calculator that operates on or creates table data:

_________________________________________________________________________________

5.1.1 Cartesian linear transformation (-Jx-JX)

There are in fact three different uses of the Cartesian linear transformation, each associated with specific command line options. The different manifestations result from specific properties of three kinds of data:

  1. Regular floating point coordinates
  2. Geographic coordinates
  3. Calendar time coordinates

Regular floating point coordinates

Selection of the Cartesian linear transformation with regular floating point coordinates will result in a simple linear scaling u = au+b of the input coordinates. The projection is defined by stating

If the y-scale or y-axis length is different from that of the x-axis (which is most often the case), separate the two scales (or lengths) by a slash, e.g., -Jx0.1i/0.5i or -JX8i/5i. Thus, our y = x data sets will plot as shown in Figure 5.1.


PIC

Figure 5.1: Linear transformation of Cartesian coordinates.

The complete commands given to produce this plot were

_________________________________________________________________________________

psxy -R0/100/0/10 -JX3i/1.5i -Ba20f10g10/a2f1g2WSne -Wthick,- -P -K sqrt.d > GMT_linear.ps  
psxy -R -J -St0.075i -Glightgray -W -O sqrt.d10 >> GMT_linear.ps
__________________________________________________________________________________________________

Normally, the user’s x-values will increase to the right and the y-values will increase upwards. It should be noted that in many situations it is desirable to have the direction of positive coordinates be reversed. For example, when plotting depth on the y-axis it makes more sense to have the positive direction downwards. All that is required to reverse the sense of positive direction is to supply a negative scale (or axis length). Finally, sometimes it is convenient to specify the width (or height) of a map and let the other dimension be computed based on the implied scale and the range of the other axis. To do this, simply specify the length to be recomputed as 0.

Geographic coordinates

PIC

Figure 5.2: Linear transformation of map coordinates.

While the Cartesian linear projection is primarily designed for regular floating point x,y data, it is sometimes necessary to plot geographical data in a linear projection. This poses a problem since longitudes have a 360° periodicity. GMT therefore needs to be informed that it has been given geographical coordinates even though a linear transformation has been chosen. We do so by adding a g (for geographical) or d (for degrees) directly after -R or by appending a g or d to the end of the -Jx (or -JX) option. As an example, we want to plot a crude world map centered on 125°E. Our command will be

_________________________________________________________________________________

gmtset GRID_CROSS_SIZE_PRIMARY 0.1i BASEMAP_TYPE FANCY PLOT_DEGREE_FORMAT ddd:mm:ssF  
pscoast -Rg-55/305/-90/90 -Jx0.014i -B60g30f15/30g30f15WSen -Dc -A1000 -Glightgray -Wthinnest -P \  
        > GMT_linear_d.ps
__________________________________________________________________________________________________

with the result reproduced in Figure 5.2.

Calendar time coordinates

PIC

Figure 5.3: Linear transformation of calendar coordinates.

Several particular issues arise when we seek to make linear plots using calendar date/time as the input coordinates. As far as setting up the coordinate transformation we must indicate whether our input data have absolute time coordinates or relative time coordinates. For the former we append T after the axis scale (or width), while for the latter we append t at the end of the -Jx (or -JX) option. However, other command line arguments (like the -R option) may already specify whether the time coordinate is absolute or relative. An absolute time entry must be given as [date]T[clock] (with date given as yyyy[-mm[-dd]], yyyy[-jjj], or yyyy[-Www[-d]], and clock using the hh[:mm[:ss[.xxx]]] 24-hour clock format) whereas the relative time is simply given as the units of time since the epoch followed by t (see TIME_UNIT and TIME_EPOCH for information on specifying the time unit and the epoch). As a simple example, we will make a plot of a school week calendar (Figure 5.3).

_________________________________________________________________________________

gmtset PLOT_DATE_FORMAT o TIME_WEEK_START Sunday PLOT_CLOCK_FORMAT -hham TIME_FORMAT_PRIMARY full  
psbasemap -R2001-9-24T/2001-09-29T/T06:59:59/T15:00:01 -JX4i/-2i -Ba1Kf1kg1d/a1Hg1hWsNe -P > GMT_linear_cal.ps
__________________________________________________________________________________________________

When the coordinate ranges provided by the -R option and the projection type given by -JX (including the optional d, g, t or T) conflict, GMT will warn the users about it. In general, the options provided with -JX will prevail.

5.1.2 Cartesian logarithmic projection


PIC

Figure 5.4: Logarithmic transformation of x-coordinates.

The log10 transformation is simply u = alog10(u)+b and is selected by appending an l (lower case L) immediately following the scale (or axis length) value. Hence, to produce a plot in which the x-axis is logarithmic (the y-axis remains linear, i.e., a semi-log plot), try

_________________________________________________________________________________

psxy -R1/100/0/10 -Jx1.5il/0.15i -B2g3/a2f1g2WSne -Wthick,- -P -K -H sqrt.d > GMT_log.ps  
psxy -R -J -Ss0.075i -Gblack -W -O -H sqrt.d10 >> GMT_log.ps
__________________________________________________________________________________________________

Note that if x- and y-scaling are different and a log10-log10 plot is desired, the l must be appended twice: Once after the x-scale (before the /) and once after the y-scale.

5.1.3 Cartesian power projection


PIC

Figure 5.5: Exponential or power transformation of x-coordinates.

This projection uses u = aub+c and allows us to explore exponential relationships like xp versus yq. While p and q can be any values, we will select p = 0.5 and q = 1 which means we will plot x versus x. We indicate this scaling by appending a p (lower case P) followed by the desired exponent, in our case 0.5. Since q = 1 we do not need to specify p1 since it is identical to the linear transformation. Thus our command becomes

_________________________________________________________________________________

psxy -R0/100/0/10 -Jx0.3ip0.5/0.15i -Ba1p/a2f1WSne -Wthick -P -K sqrt.d > GMT_pow.ps  
psxy -R -J -Sc0.075i -Gwhite -W -O sqrt.d10 >> GMT_pow.ps
_________________________________________________________________________________________________

5.2 Linear projection with polar (θ,r) coordinates (-Jp -JP)


PIC

Figure 5.6: Polar (Cylindrical) transformation of (θ,r) coordinates.

This transformation converts polar coordinates (angle θ and radius r) to positions on a plot. Now x = f(θ,r) and y = g(θ,r), hence it is similar to a regular map projection because x and y are coupled and x (i.e., θ) has a 360° periodicity. With input and output points both in the plane it is a two-dimensional projection. The transformation comes in two flavors:

  1. Normally, θ is understood to be directions counter-clockwise from the horizontal axis, but we may choose to specify an angular offset [whose default value is zero]. We will call this offset θ0. Then, x = f(θ,r) = arcos(θ -θ0)+b and y = g(θ,r) = arsin(θ -θ0)+c.
  2. Alternatively, θ can be interpreted to be azimuths clockwise from the vertical axis, yet we may again choose to specify the angular offset [whose default value is zero]. Then, x = f(θ,r) = arcos(90-(θ -θ0))+b and y = g(θ,r) = arsin(90-(θ -θ0))+c.

Consequently, the polar transformation is defined by providing

As an example of this projection we will create a gridded data set in polar coordinates z(θ,r) = r2 cos4θ using grdmath, a RPN calculator that operates on or creates grid files.

_________________________________________________________________________________

grdmath -R0/360/2/4 -I6/0.1 X 4 MUL PI MUL 180 DIV COS Y 2 POW MUL = $$.nc  
grdcontour $$.nc -JP3i -B30Ns -P -C2 -S4 --PLOT_DEGREE_FORMAT=+ddd > GMT_polar.ps  
rm -f $$.nc
__________________________________________________________________________________________________

We used grdcontour to make a contour map of this data. Because the data file only contains values with 2 r 4, a donut shaped plot appears in Figure 5.6

Chapter 6
GMT Map Projections

GMT implements more than 30 different projections. They all project the input coordinates longitude and latitude to positions on a map. In general, x = f(x,y,z) and y = g(x,y,z), where z is implicitly given as the radial vector length to the (x,y) point on the chosen ellipsoid. The functions f and g can be quite nasty and we will refrain from presenting details in this document. The interested read is referred to Snyder [1987]21. We will mostly be using the pscoast command to demonstrate each of the projections. GMT map projections are grouped into four categories depending on the nature of the projection. The groups are

  1. Conic map projections
  2. Azimuthal map projections
  3. Cylindrical map projections
  4. Miscellaneous projections

Because x and y are coupled we can only specify one plot-dimensional scale, typically a map scale (for lower-case map projection code) or a map width (for upper-case map projection code). However, in some cases it would be more practical to specify map height instead of width, while in other situations it would be nice to set either the shortest or longest map dimension. Users may select these alternatives by appending a character code to their map dimension. To specify map height, append h to the given dimension; to select the minimum map dimension, append -, whereas you may append + to select the maximum map dimension. Without the modifier the map width is selected by default.

In GMT version 4.3.0 we noticed we ran out of the alphabet for 1-letter (and sometimes 2-letter) projection codes. To allow more flexibility, and to make it easier to remember the codes, we implemented the option to use the abbreviations used by the Proj4 mapping package. Since some of the GMT projections are not in Proj4, we invented some of our own as well. For a full list of both the old 1- and 2-letter codes, as well as the Proj4-equivalents see the quick reference cards in Section 3.2. For example, -JM15c and -JMerc/15c have the same meaning.

6.1 Conic projections

6.1.1 Albers conic equal-area projection (-Jb-JB)

This projection, developed by Albers in 1805, is predominantly used to map regions of large east-west extent, in particular the United States. It is a conic, equal-area projection, in which parallels are unequally spaced arcs of concentric circles, more closely spaced at the north and south edges of the map. Meridians, on the other hand, are equally spaced radii about a common center, and cut the parallels at right angles. Distortion in scale and shape vanishes along the two standard parallels. Between them, the scale along parallels is too small; beyond them it is too large. The opposite is true for the scale along meridians. To define the projection in GMT you need to provide the following information:

Note that you must include the “1:” if you choose to specify the scale that way. E.g., you can say 0.5 which means 0.5 inch/degree or 1:200000 which means 1 inch on the map equals 200,000 inches along the standard parallels. The projection center defines the origin of the rectangular map coordinates. As an example we will make a map of the region near Taiwan. We choose the center of the projection to be at 125 °E/20 °N and 25 °N and 45 °N as our two standard parallels. We desire a map that is 5 inches wide. The complete command needed to generate the map below is therefore given by:

_________________________________________________________________________________

gmtset GRID_CROSS_SIZE_PRIMARY 0  
pscoast -R110/140/20/35 -JB125/20/25/45/5i -B10g5 -Dl -Glightgray -Wthinnest -A250 -P > GMT_albers.ps
__________________________________________________________________________________________________

PIC

Figure 6.1: Albers equal-area conic map projection

6.1.2 Equidistant conic projection (-Jd-JD)

The equidistant conic projection was described by the Greek philosopher Claudius Ptolemy about A.D. 150. It is neither conformal or equal-area, but serves as a compromise between them. The scale is true along all meridians and the standard parallels. To select this projection in GMT you must provide the same information as for the other conic projection, i.e.,

The equidistant conic projection is often used for atlases with maps of small countries. As an example, we generate a map of Cuba:

_________________________________________________________________________________

gmtset PLOT_DEGREE_FORMAT ddd:mm:ssF GRID_CROSS_SIZE_PRIMARY 0.05i  
pscoast -R-88/-70/18/24 -JD-79/21/19/23/4.5i -B5g1 -Di -N1/thick -Glightgray \  
        -Wthinnest -P > GMT_equidistant_conic.ps
__________________________________________________________________________________________________

PIC

Figure 6.2: Equidistant conic map projection

6.1.3 Lambert conic conformal projection (-Jl-JL)

This conic projection was designed by the Alsatian mathematician Johann Heinrich Lambert (1772) and has been used extensively for mapping of regions with predominantly east-west orientation, just like the Albers projection. Unlike the Albers projection, Lambert’s conformal projection is not equal-area. The parallels are arcs of circles with a common origin, and meridians are the equally spaced radii of these circles. As with Albers projection, it is only the two standard parallels that are distortion-free. To select this projection in GMT you must provide the same information as for the Albers projection, i.e.,

The Lambert conformal projection has been used for basemaps for all the 48 contiguous States with the two fixed standard parallels 33°N and 45°N. We will generate a map of the continental USA using these parameters. Note that with all the projections you have the option of selecting a rectangular border rather than one defined by meridians and parallels. Here, we choose the regular WESN region, a “fancy” basemap frame, and use degrees west for longitudes. The generating commands used were

_________________________________________________________________________________

gmtset BASEMAP_TYPE FANCY PLOT_DEGREE_FORMAT ddd:mm:ssF GRID_CROSS_SIZE_PRIMARY 0.05i  
pscoast -R-130/-70/24/52 -Jl-100/35/33/45/1:50000000 -B10g5 -Dl -N1/thick -N2/thinner -A500 \  
        -Glightgray -Wthinnest -P > GMT_lambert_conic.ps
__________________________________________________________________________________________________

PIC

Figure 6.3: Lambert conformal conic map projection

The choice for projection center does not affect the projection but it indicates which meridian (here 100°W) will be vertical on the map. The standard parallels were originally selected by Adams to provide a maximum scale error between latitudes 30.5°N and 47.5°N of 0.5–1%. Some areas, like Florida, experience scale errors of up to 2.5%.

6.1.4 (American) polyconic projection (-Jpoly-JPoly

The polyconic projection, in Europe usually referred to as the American polyconic projection, was introduced shortly before 1820 by the Swiss-American cartographer Ferdinand Rodulph Hassler (1770-1843). As head of the Survey of the Coast, he was looking for a projection that would give the least distortion for mapping the coast of the United States. The projection acquired its name from the construction of each parallel, which is achieved by projecting the parallel onto the cone while it is rolled around the globe, along the central meridian, tangent to that parallel. As a consequence, the projection involves many cones rather than a single one used in regular conic projections.

The polyconic projection is neither equal-area, nor conformal. It is true to scale without distortion along the central meridian. Each parallel is true to scale as well, but the meridians are not as they get further away from the central meridian. As a consequence, no parallel is standard because conformity is lost with the lengthening of the meridians.

Below we reproduce the illustration by Snyder [1987], with a gridline every 10° and annotations only every 30° in longitude:

_________________________________________________________________________________

pscoast -R-180/-20/0/90 -JPoly/4i -B30g10/10g10 -Dc -A1000 -Glightgray -Wthinnest -P \  
        > GMT_polyconic.ps
__________________________________________________________________________________________________

PIC

Figure 6.4: (American) polyconic projection

6.2 Azimuthal projections

6.2.1 Lambert Azimuthal Equal-Area (-Ja-JA)

This projection was developed by Lambert in 1772 and is typically used for mapping large regions like continents and hemispheres. It is an azimuthal, equal-area projection, but is not perspective. Distortion is zero at the center of the projection, and increases radially away from this point. To define this projection in GMT you must provide the following information:

Two different types of maps can be made with this projection depending on how the region is specified. We will give examples of both types.

Rectangular map

In this mode we define our region by specifying the longitude/latitude of the lower left and upper right corners instead of the usual west, east, south, north boundaries. The reason for specifying our area this way is that for this and many other projections, lines of equal longitude and latitude are not straight lines and are thus poor choices for map boundaries. Instead we require that the map boundaries be rectangular by defining the corners of a rectangular map boundary. Using 0°E/40°S (lower left) and 60°E/10°S (upper right) as our corners we try

_________________________________________________________________________________

gmtset PLOT_DEGREE_FORMAT ddd:mm:ssF GRID_CROSS_SIZE_PRIMARY 0  
pscoast -R0/-40/60/-10r -JA30/-30/4.5i -B30g30/15g15 -Dl -A500 -Glightgray -Wthinnest -P \  
        > GMT_lambert_az_rect.ps
__________________________________________________________________________________________________

PIC

Figure 6.5: Rectangular map using the Lambert azimuthal equal-area projection.

Note that an “r” is appended to the -R option to inform GMT that the region has been selected using the rectangle technique, otherwise it would try to decode the values as west, east, south, north and report an error since ’east’ < ’west’.

Hemisphere map

Here, you must specify the world as your region (-Rg or -Rd). E.g., to obtain a hemisphere view that shows the Americas, try

_________________________________________________________________________________

pscoast -Rg -JA280/30/3.5i -B30g30/15g15 -Dc -A1000 -Gblack -P > GMT_lambert_az_hemi.ps
__________________________________________________________________________________________________

PIC

Figure 6.6: Hemisphere map using the Lambert azimuthal equal-area projection.

To geologists, the Lambert azimuthal equal-area projection (with origin at 0°/0°) is known as the equal-area (Schmidt) stereonet and used for plotting fold axes, fault planes, and the like. An equal-angle (Wulff) stereonet can be obtained by using the stereographic projection (discussed later). The stereonets produced by these two projections appear below.


PIC

Figure 6.7: Equal-Area (Schmidt) and Equal-Angle (Wulff) stereo nets.

6.2.2 Stereographic Equal-Angle projection (-Js-JS)

This is a conformal, azimuthal projection that dates back to the Greeks. Its main use is for mapping the polar regions. In the polar aspect all meridians are straight lines and parallels are arcs of circles. While this is the most common use it is possible to select any point as the center of projection. The requirements are

A default map scale factor of 0.9996 will be applied by default (although you may change this with MAP_SCALE_FACTOR). However, the setting is ignored when a standard parallel has been specified since the scale is then implicitly given. We will look at two different types of maps.

Polar Stereographic Map

In our first example we will let the projection center be at the north pole. This means we have a polar stereographic projection and the map boundaries will coincide with lines of constant longitude and latitude. An example is given by

_________________________________________________________________________________

pscoast -R-30/30/60/72 -Js0/90/4.5i/60 -Ba10g5/5g5 -Dl -A250 -Gblack -P > GMT_stereographic_polar.ps
__________________________________________________________________________________________________

PIC

Figure 6.8: Polar stereographic conformal projection.

Rectangular stereographic map

As with Lambert’s azimuthal equal-area projection we have the option to use rectangular boundaries rather than the wedge-shape typically associated with polar projections. This choice is defined by selecting two points as corners in the rectangle and appending an “r” to the -R option. This command produces a map as presented in Figure 6.9:

_________________________________________________________________________________

gmtset OBLIQUE_ANNOTATION 30  
pscoast -R-25/59/70/72r -JS10/90/11c -B30g10/5g5 -Dl -A250 -Glightgray -Wthinnest -P \  
        > GMT_stereographic_rect.ps
__________________________________________________________________________________________________

PIC

Figure 6.9: Polar stereographic conformal projection with rectangular borders.

General stereographic map

In terms of usage this projection is identical to the Lambert azimuthal equal-area projection. Thus, one can make both rectangular and hemispheric maps. Our example shows Australia using a projection pole at 130E/30°S. The command used was

_________________________________________________________________________________

gmtset OBLIQUE_ANNOTATION 0  
pscoast -R100/-40/160/-10r -JS130/-30/4i -B30g10/15g15 -Dl -A500 -Gblack -P \  
        > GMT_stereographic_general.ps
__________________________________________________________________________________________________

PIC

Figure 6.10: General stereographic conformal projection with rectangular borders.

By choosing 0°/0°as the pole, we obtain the conformal stereonet presented next to its equal-area cousin in the Section 6.2.1 on the Lambert azimuthal equal-area projection (Figure 6.7).

6.2.3 Perspective projection (-Jg-JG)

The perspective projection imitates in 2 dimensions the 3-dimensional view of the earth from space. The implementation in GMT is very flexible, and thus requires many input variables. Those are listed and explained below, with the values used in Figure 6.11 between brackets.

The imagined view of northwest Europe from a Space Shuttle at 230 km looking due east is thus accomplished by the following pscoast command:

_________________________________________________________________________________

pscoast -Rg -JG4/52/230/90/60/180/60/60/5i -B2g2/1g1 -Ia -Di -Glightgray -Wthinnest -P \  
        --ANNOT_MIN_SPACING=0.25i > GMT_perspective.ps
__________________________________________________________________________________________________

PIC

Figure 6.11: View from the Space Shuttle in Perspective projection.

6.2.4 Orthographic projection (-Jg-JG)

The orthographic azimuthal projection is a perspective projection from infinite distance. It is therefore often used to give the appearance of a globe viewed from outer space. As with Lambert’s equal-area and the stereographic projection, only one hemisphere can be viewed at any time. The projection is neither equal-area nor conformal, and much distortion is introduced near the edge of the hemisphere. The directions from the center of projection are true. The projection was known to the Egyptians and Greeks more than 2,000 years ago. Because it is mainly used for pictorial views at a small scale, only the spherical form is necessary.

To specify the orthographic projection the same options -Jg or -JG as the perspective projection are used, but with fewer variables to supply:

Our example of a perspective view centered on 75°W/40°N can therefore be generated by the following pscoast command:

_________________________________________________________________________________

pscoast -Rg -JG-75/41/4.5i -B15g15 -Dc -A5000 -Gblack -P > GMT_orthographic.ps
__________________________________________________________________________________________________

PIC

Figure 6.12: Hemisphere map using the Orthographic projection.

6.2.5 Azimuthal Equidistant projection (-Je-JE)

The most noticeable feature of this azimuthal projection is the fact that distances measured from the center are true. Therefore, a circle about the projection center defines the locus of points that are equally far away from the plot origin. Furthermore, directions from the center are also true. The projection, in the polar aspect, is at least several centuries old. It is a useful projection for a global view of locations at various or identical distance from a given point (the map center).

To specify the azimuthal equidistant projection you must supply:

Our example of a global view centered on 100°W/40°N can therefore be generated by the following pscoast command. Note that the antipodal point is 180° away from the center, but in this projection this point plots as the entire map perimeter:

_________________________________________________________________________________

pscoast -Rg -JE-100/40/4.5i -B15g15 -Dc -A10000 -Glightgray -Wthinnest -P > GMT_az_equidistant.ps
__________________________________________________________________________________________________

PIC

Figure 6.13: World map using the equidistant azimuthal projection.

6.2.6 Gnomonic projection (-Jf-JF)

The Gnomonic azimuthal projection is a perspective projection from the center onto a plane tangent to the surface. Its origin goes back to the old Greeks who used it for star maps almost 2500 years ago. The projection is neither equal-area nor conformal, and much distortion is introduced near the edge of the hemisphere; in fact, less than a hemisphere may be shown around a given center. The directions from the center of projection are true. Great circles project onto straight lines. Because it is mainly used for pictorial views at a small scale, only the spherical form is necessary.

To specify the Gnomonic projection you must supply:

Using a horizon of 60°, our example of this projection centered on 120°W/35°N can therefore be generated by the following pscoast command:

_________________________________________________________________________________

pscoast -Rg -JF-120/35/60/4.5i -B30g15 -Dc -A10000 -Glightgray -Wthinnest -P > GMT_gnomonic.ps
__________________________________________________________________________________________________

PIC

Figure 6.14: Gnomonic azimuthal projection.

6.3 Cylindrical projections

Cylindrical projections are easily recognized for its shape: maps are rectangular and meridians and parallels are straight lines crossing at right angles. But that is where similarities between the cylindrical projections supported by GMT (Mercator, transverse Mercator, universal transverse Mercator, oblique Mercator, Cassini, cylindrical equidistant, cylindrical equal-area, Miller, and cylindrical stereographic projections) stops. Each have a different way of spacing the meridians and parallels to obtain certain desirable cartographic properties.

6.3.1 Mercator projection (-Jm-JM)

Probably the most famous of the various map projections, the Mercator projection takes its name from the Flemish cartographer Gheert Cremer, better known as Gerardus Mercator, who presented it in 1569. The projection is a cylindrical and conformal, with no distortion along the equator. A major navigational feature of the projection is that a line of constant azimuth is straight. Such a line is called a rhumb line or loxodrome. Thus, to sail from one point to another one only had to connect the points with a straight line, determine the azimuth of the line, and keep this constant course for the entire voyage22. The Mercator projection has been used extensively for world maps in which the distortion towards the polar regions grows rather large, thus incorrectly giving the impression that, for example, Greenland is larger than South America. In reality, the latter is about eight times the size of Greenland. Also, the Former Soviet Union looks much bigger than Africa or South America. One may wonder whether this illusion has had any influence on U.S. foreign policy.

In the regular Mercator projection, the cylinder touches the globe along the equator. Other orientations like vertical and oblique give rise to the Transverse and Oblique Mercator projections, respectively. We will discuss these generalizations following the regular Mercator projection.

The regular Mercator projection requires a minimum of parameters. To use it in GMT programs you supply this information (the first two items are optional and have defaults):

Our example presents a world map at a scale of 0.012 inch pr degree which will give a map 4.32 inch wide. It was created with the command:

_________________________________________________________________________________

gmtset BASEMAP_TYPE fancy  
pscoast -R0/360/-70/70 -Jm1.2e-2i -Ba60f30/a30f15 -Dc -A5000 -Gblack -P > GMT_mercator.ps
__________________________________________________________________________________________________

PIC

Figure 6.15: Simple Mercator map.

While this example is centered on the Dateline, one can easily choose another configuration with the -R option. A map centered on Greenwich would specify the region with -R-180/180/-70/70.

6.3.2 Transverse Mercator projection (-Jt-JT)

The transverse Mercator was invented by Lambert in 1772. In this projection the cylinder touches a meridian along which there is no distortion. The distortion increases away from the central meridian and goes to infinity at 90° from center. The central meridian, each meridian 90° away from the center, and equator are straight lines; other parallels and meridians are complex curves. The projection is defined by specifying:

The optional latitude of origin defaults to Equator if not specified. Although defaulting to 1, you can change the map scale factor via the MAP_SCALE_FACTOR parameter. Our example shows a transverse Mercator map of south-east Europe and the Middle East with 35°E as the central meridian:

_________________________________________________________________________________

pscoast -R20/30/50/45r -Jt35/0.18i -B10g5 -Dl -A250 -Glightgray -Wthinnest -P \  
        > GMT_transverse_merc.ps
__________________________________________________________________________________________________

PIC

Figure 6.16: Rectangular Transverse Mercator map.

The transverse Mercator can also be used to generate a global map—the equivalent of the 360° Mercator map. Using the command

_________________________________________________________________________________

pscoast -R0/360/-80/80 -JT330/-45/3.5i -B30g15/15g15WSne -Dc -A2000 -Gblack -P > GMT_TM.ps
__________________________________________________________________________________________________

we made the map illustrated in Figure 6.17. Note that when a world map is given (indicated by -R0/360/s/n), the arguments are interpreted to mean oblique degrees, i.e., the 360° range is understood to mean the extent of the plot along the central meridian, while the “south” and “north” values represent how far from the central longitude we want the plot to extend. These values correspond to latitudes in the regular Mercator projection and must therefore be less than 90°.


PIC

Figure 6.17: A global transverse Mercator map.

6.3.3 Universal Transverse Mercator (UTM) projection (-Ju-JU)

A particular subset of the transverse Mercator is the Universal Transverse Mercator (UTM) which was adopted by the US Army for large-scale military maps. Here, the globe is divided into 60 zones between 84°S and 84°N, most of which are 6° wide. Each of these UTM zones have their unique central meridian. Furthermore, each zone is divided into latitude bands but these are not needed to specify the projection for most cases. See Figure 6.18 for all zone designations.


PIC

Figure 6.18: Universal Transverse Mercator zone layout.

GMT implements both the transverse Mercator and the UTM projection. When selecting UTM you must specify:

In order to minimize the distortion in any given zone, a scale factor of 0.9996 has been factored into the formulae. (although a standard, you can change this with MAP_SCALE_FACTOR). This makes the UTM projection a secant projection and not a tangent projection like the transverse Mercator above. The scale only varies by 1 part in 1,000 from true scale at equator. The ellipsoidal projection expressions are accurate for map areas that extend less than 10° away from the central meridian. For larger regions we use the conformal latitude in the general spherical formulae instead.

6.3.4 Oblique Mercator projection (-Jo-JO)

Oblique configurations of the cylinder give rise to the oblique Mercator projection. It is particularly useful when mapping regions of large lateral extent in an oblique direction. Both parallels and meridians are complex curves. The projection was developed in the early 1900s by several workers. Several parameters must be provided to define the projection. GMT offers three different definitions:

  1. Option -Joa or -JOa:
  2. Option -Job or -JOb:
  3. Option -Joc or -JOc:

Our example was produced by the command

_________________________________________________________________________________

pscoast -R270/20/305/25r -JOc280/25.5/22/69/4.8i -B10g5 -Di -A250 -Glightgray -Wthinnest -P \  
        -Tf301.5/23/0.4i/2 --HEADER_FONT_SIZE=8p --HEADER_OFFSET=0.05i > GMT_obl_merc.ps
__________________________________________________________________________________________________

PIC

Figure 6.19: Oblique Mercator map using -Joc. We make it clear which direction is North by adding a star rose with the -T option.

It uses definition 3 for an oblique view of some Caribbean islands. Note that we define our region using the rectangular system described earlier. If we do not append an “r” to the -R string then the information provided with the -R option is assumed to be oblique degrees about the projection center rather than the usual geographic coordinates. This interpretation is chosen since in general the parallels and meridians are not very suitable as map boundaries.

6.3.5 Cassini cylindrical projection (-Jc-JC)

This cylindrical projection was developed in 1745 by César-François Cassini de Thury for the survey of France. It is occasionally called Cassini-Soldner since the latter provided the more accurate mathematical analysis that led to the development of the ellipsoidal formulae. The projection is neither conformal nor equal-area, and behaves as a compromise between the two end-members. The distortion is zero along the central meridian. It is best suited for mapping regions of north-south extent. The central meridian, each meridian 90° away, and equator are straight lines; all other meridians and parallels are complex curves. The requirements to define this projection are:

A detailed map of the island of Sardinia centered on the 8°45’E meridian using the Cassini projection can be obtained by running the command:

_________________________________________________________________________________

pscoast -R7:30/38:30/10:30/41:30r -JC8.75/40/2.5i -B1g1f30m -Lf9.5/38.8/40/60 -Dh -Glightgray \  
        -Wthinnest -Ia/thinner -P --LABEL_FONT_SIZE=12 > GMT_cassini.ps
__________________________________________________________________________________________________

PIC

Figure 6.20: Cassini map over Sardinia.

As with the previous projections, the user can choose between a rectangular boundary (used here) or a geographical (WESN) boundary.

6.3.6 Cylindrical equidistant projection (-Jq-JQ)

This simple cylindrical projection is really a linear scaling of longitudes and latitudes. The most common form is the Plate Carrée projection, where the scaling of longitudes and latitudes is the same. All meridians and parallels are straight lines. The projection can be defined by:

The first two of these are optional and have defaults. When the standard parallel is defined, the central meridian must be supplied as well.

A world map centered on the dateline using this projection can be obtained by running the command:

_________________________________________________________________________________

pscoast -Rg -JQ4.5i -B60f30g30 -Dc -A5000 -Gblack -P > GMT_equi_cyl.ps
__________________________________________________________________________________________________

PIC

Figure 6.21: World map using the Plate Carrée projection.

Different relative scalings of longitudes and latitudes can be obtained by selecting a standard parallel different from the equator. Some selections for standard parallels have practical properties as shown in Table 6.1.




Projection
Standard parallel


Grafarend and Niermann, minimum linear distortion 61.7°
Ronald Miller Equirectangular 50.5°
Ronald Miller, minimum continental distortion 43.5°
Grafarend and Niermann 42°
Ronald Miller, minimum overall distortion 37.5°
Plate Carrée, Simple Cylindrical, Plain/Plane 0°



Table 6.1: Standard parallels for some cylindrical equidistant projections.

6.3.7 Cylindrical equal-area projections (-Jy-JY)

This cylindrical projection is actually several projections, depending on what latitude is selected as the standard parallel. However, they are all equal area and hence non-conformal. All meridians and parallels are straight lines. The requirements to define this projection are:

While you may choose any value for the standard parallel and obtain your own personal projection, there are seven choices of standard parallels that result in known (or named) projections. These are listed in Table 6.2.




Projection
Standard parallel


Balthasart 50°
Gall 45°
Hobo-Dyer 37°30’ (= 37.5°)
Trystan Edwards 37°24’ (= 37.4°)
Caster 37°04’ (= 37.0666°)
Behrman 30°
Lambert 0°



Table 6.2: Standard parallels for some cylindrical equal-area projections.

For instance, a world map centered on the 35°E meridian using the Behrman projection (Figure 6.22) can be obtained by running the command:

_________________________________________________________________________________

pscoast -R-145/215/-90/90 -JY35/30/4.5i -B45g45 -Dc -A10000 -Slightgray -Wthinnest -P > \  
        GMT_general_cyl.ps
__________________________________________________________________________________________________

PIC

Figure 6.22: World map using the Behrman cylindrical equal-area projection.

As one can see there is considerable distortion at high latitudes since the poles map into lines.

6.3.8 Miller Cylindrical projection (-Jj-JJ)

This cylindrical projection, presented by Osborn Maitland Miller of the American Geographic Society in 1942, is neither equal nor conformal. All meridians and parallels are straight lines. The projection was designed to be a compromise between Mercator and other cylindrical projections. Specifically, Miller spaced the parallels by using Mercator’s formula with 0.8 times the actual latitude, thus avoiding the singular poles; the result was then divided by 0.8. There is only a spherical form for this projection. Specify the projection by:

For instance, a world map centered on the 90°E meridian at a map scale of 1:400,000,000 (Figure 6.23) can be obtained as follows:

_________________________________________________________________________________

pscoast -R-90/270/-80/90 -Jj1:400000000 -B45g45/30g30 -Dc -A10000 -Glightgray -Wthinnest -P \  
        > GMT_miller.ps
__________________________________________________________________________________________________

PIC

Figure 6.23: World map using the Miller cylindrical projection.

6.3.9 Cylindrical stereographic projections (-Jcyl_stere-JCyl_stere)

The cylindrical stereographic projections are certainly not as notable as other cylindrical projections, but are still used because of their relative simplicity and their ability to overcome some of the downsides of other cylindrical projections, like extreme distortions of the higher latitudes. The stereographic projections are perspective projections, projecting the sphere onto a cylinder in the direction of the antipodal point on the equator. The cylinder crosses the sphere at two standard parallels, equidistant from the equator. The projections are defined by:

Some of the selections of the standard parallel are named for the cartographer or publication that popularized the projection (Table 6.3).




Projection
Standard parallel


Miller’s modified Gall 66.159467°
Kamenetskiy’s First 55°
Gall’s stereographic 45°
Bolshoi Sovietskii Atlas Mira or Kamenetskiy’s Second 30°
Braun’s cylindrical 0°



Table 6.3: Standard parallels for some cylindrical equal-area projections.

A map of the world, centered on the Greenwich meridian, using the Gall’s stereographic projection (standard parallel is 45°, Figure 6.24), is obtained as follows:

_________________________________________________________________________________

gmtset PLOT_DEGREE_FORMAT dddA  
pscoast -R-180/180/-60/80 -JCyl_stere/0/45/4.5i -Ba60f30g30/a30g30 -Dc -A5000 -Wblack -Ggrey -P \  
        > GMT_gall_stereo.ps
__________________________________________________________________________________________________

PIC

Figure 6.24: World map using Gall’s stereographic projection.

6.4 Miscellaneous projections

GMT supports 8 common projections for global presentation of data or models. These are the Hammer, Mollweide, Winkel Tripel, Robinson, Eckert IV and VI, Sinusoidal, and Van der Grinten projections. Due to the small scale used for global maps these projections all use the spherical approximation rather than more elaborate elliptical formulae.

In all cases, the specification of the central meridian can be skipped. The default is the middle of the longitude range of the plot, specified by the -(R) option.

6.4.1 Hammer projection (-Jh-JH)

The equal-area Hammer projection, first presented by the German mathematician Ernst von Hammer in 1892, is also known as Hammer-Aitoff (the Aitoff projection looks similar, but is not equal-area). The border is an ellipse, equator and central meridian are straight lines, while other parallels and meridians are complex curves. The projection is defined by selecting:

A view of the Pacific ocean using the Dateline as central meridian is accomplished thus

_________________________________________________________________________________

pscoast -Rg -JH4.5i -Bg30/g15 -Dc -A10000 -Gblack -P > GMT_hammer.ps
__________________________________________________________________________________________________

PIC

Figure 6.25: World map using the Hammer projection.

6.4.2 Mollweide projection (-Jw-JW)

This pseudo-cylindrical, equal-area projection was developed by the German mathematician and astronomer Karl Brandan Mollweide in 1805. Parallels are unequally spaced straight lines with the meridians being equally spaced elliptical arcs. The scale is only true along latitudes 40°44’ north and south. The projection is used mainly for global maps showing data distributions. It is occasionally referenced under the name homalographic projection. Like the Hammer projection, outlined above, we need to specify only two parameters to completely define the mapping of longitudes and latitudes into rectangular x/y coordinates:

An example centered on Greenwich can be generated thus:

_________________________________________________________________________________

pscoast -Rd -JW4.5i -Bg30/g15 -Dc -A10000 -Gblack -P > GMT_mollweide.ps
__________________________________________________________________________________________________

PIC

Figure 6.26: World map using the Mollweide projection.

6.4.3 Winkel Tripel projection (-Jr-JR)

In 1921, the German mathematician Oswald Winkel a projection that was to strike a compromise between the properties of three elements (area, angle and distance). The German word “tripel” refers to this junction of where each of these elements are least distorted when plotting global maps. The projection was popularized when Bartholomew and Son started to use it in its world-renowned “The Times Atlas of the World” in the mid 20th century. In 1998, the National Geographic Society made the Winkel Tripel as its map projection of choice for global maps.

Naturally, this projection is neither conformal, nor equal-area. Central meridian and equator are straight lines; other parallels and meridians are curved. The projection is obtained by averaging the coordinates of the Equidistant Cylindrical and Aitoff (not Hammer-Aitoff) projections. The poles map into straight lines 0.4 times the length of equator. To use it you must enter

Centered on Greenwich, the example in Figure 6.27 was created by this command:

_________________________________________________________________________________

pscoast -Rd -JR4.5i -Bg30/g15 -Dc -A10000 -Ggray -P > GMT_winkel.ps
__________________________________________________________________________________________________

PIC

Figure 6.27: World map using the Winkel Tripel projection.

6.4.4 Robinson projection (-Jn-JN)

The Robinson projection, presented by the American geographer and cartographer Arthur H. Robinson in 1963, is a modified cylindrical projection that is neither conformal nor equal-area. Central meridian and all parallels are straight lines; other meridians are curved. It uses lookup tables rather than analytic expressions to make the world map “look” right23. The scale is true along latitudes ±38°. The projection was originally developed for use by Rand McNally and is currently used by the National Geographic Society. To use it you must enter

Again centered on Greenwich, the example below was created by this command:

_________________________________________________________________________________

pscoast -Rd -JN4.5i -Bg30/g15 -Dc -A10000 -Ggray -P > GMT_robinson.ps
__________________________________________________________________________________________________

PIC

Figure 6.28: World map using the Robinson projection.

6.4.5 Eckert IV and VI projection (-Jk-JK)

The Eckert IV and VI projections, presented by the German cartographer Max Eckert-Greiffendorff in 1906, are pseudocylindrical equal-area projections. Central meridian and all parallels are straight lines; other meridians are equally spaced elliptical arcs (IV) or sinusoids (VI). The scale is true along latitudes ±40°30’ (IV) and ±49°16’ (VI). Their main use is in thematic world maps. To select Eckert IV you must use -JKf (f for “four”) while Eckert VI is selected with -JKs (s for “six”). If no modifier is given it defaults to Eckert VI. In addition, you must enter

Centered on the Dateline, the Eckert IV example below was created by this command:

_________________________________________________________________________________

pscoast -Rg -JKf4.5i -Bg30/g15 -Dc -A10000 -Wthinnest -Gwhite -Slightgray -P > GMT_eckert4.ps
__________________________________________________________________________________________________

PIC

Figure 6.29: World map using the Eckert IV projection.

The same script, with s instead of f, yields the Eckert VI map:


PIC

Figure 6.30: World map using the Eckert VI projection.

6.4.6 Sinusoidal projection (-Ji-JI)

The sinusoidal projection is one of the oldest known projections, is equal-area, and has been used since the mid-16th century. It has also been called the “Equal-area Mercator” projection. The central meridian is a straight line; all other meridians are sinusoidal curves. Parallels are all equally spaced straight lines, with scale being true along all parallels (and central meridian). To use it, you need to select:

A simple world map using the sinusoidal projection is therefore obtained by

_________________________________________________________________________________

pscoast -Rd -JI4.5i -Bg30/g15 -Dc -A10000 -Ggray -P > GMT_sinusoidal.ps
__________________________________________________________________________________________________

PIC

Figure 6.31: World map using the Sinusoidal projection.

To reduce distortion of shape the interrupted sinusoidal projection was introduced in 1927. Here, three symmetrical segments are used to cover the entire world. Traditionally, the interruptions are at 160°W, 20°W, and 60°E. To make the interrupted map we must call pscoast for each segment and superpose the results. To produce an interrupted world map (with the traditional boundaries just mentioned) that is 5.04 inches wide we use the scale 5.04/360° = 0.014 and offset the subsequent plots horizontally by their widths (140°0.014 and 80°0.014):

_________________________________________________________________________________

pscoast -R200/340/-90/90 -Ji0.014i -Bg30/g15 -A10000 -Dc -Gblack -K -P > GMT_sinus_int.ps  
pscoast -R-20/60/-90/90 -Ji0.014i -Bg30/g15 -Dc -A10000 -Gblack -X1.96i -O -K >> GMT_sinus_int.ps  
pscoast -R60/200/-90/90 -Ji0.014i -Bg30/g15 -Dc -A10000 -Gblack -X1.12i -O >> GMT_sinus_int.ps
__________________________________________________________________________________________________

PIC

Figure 6.32: World map using the Interrupted Sinusoidal projection.

The usefulness of the interrupted sinusoidal projection is basically limited to display of global, discontinuous data distributions like hydrocarbon and mineral resources, etc.

6.4.7 Van der Grinten projection (-Jv-JV)

The Van der Grinten projection, presented by Alphons J. van der Grinten in 1904, is neither equal-area nor conformal. Central meridian and Equator are straight lines; other meridians are arcs of circles. The scale is true along the Equator only. Its main use is to show the entire world enclosed in a circle. To use it you must enter

Centered on the Dateline, the example below was created by this command:

_________________________________________________________________________________

pscoast -Rg -JV4i -Bg30/g15 -Dc -Glightgray -A10000 -Wthinnest -P > GMT_grinten.ps
__________________________________________________________________________________________________

PIC

Figure 6.33: World map using the Van der Grinten projection.

Chapter 7
Creating GMT Graphics

In this section we will be giving several examples of typical usage of GMT programs. In general, we will start with a raw data set, manipulate the numbers in various ways, then display the results in diagram or map view. The resulting plots will have in common that they are all made up of simpler plots that have been overlaid to create a complex illustration. We will mostly follow the following format:

  1. We explain what we want to achieve in plain language.
  2. We present an annotated Bourne shell script that contains all commands used to generate the illustration.
  3. We explain the rationale behind the commands.
  4. We present the illustration, 50% reduced in size, and without the timestamp (-U).

A detailed discussion of each command is not given; we refer you to the manual pages for command line syntax, etc. We encourage you to run these scripts for yourself. See Appendix D if you would like an electronic version of all the shell-scripts (both sh and csh scripts are available, as or DOS batch files; only the sh-scripts are discussed here) and support data used below. Note that all examples explicitly specifies the measurement units, so although we use inches you should be able to run these scripts and get the same plots even if you have cm as the default measure unit. The examples are all written to be “quiet”, that is no information is echoed to the screen. Thus, these scripts are well suited for background execution.

Note that we also end each script by cleaning up after ourselves. Because awk is broken as designed on some systems, and nawk is not available on others we refer to $AWK in the scripts below; the do_examples.sh scripts will set this when running all examples.

Finally, be aware that for practical purposes the output PostScript file name is stored as the variable ps.

7.1 The making of contour maps

We want to create two contour maps of the low order geoid using the Hammer equal area projection. Our gridded data file is called osu91a1f_16.nc and contains a global 1° by 1° gridded geoid (we will see how to make gridded files later). We would like to show one map centered on Greenwich and one centered on the dateline. Positive contours should be drawn with a solid pen and negative contours with a dashed pen. Annotations should occur for every 50 m contour level, and both contour maps should show the continents in light gray in the background. Finally, we want a rectangular frame surrounding the two maps. This is how it is done:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 01  
#  
# Purpose:      Make two contour maps based on the data in the file osu91a1f_16.nc  
# GMT progs:    gmtset, grdcontour, psbasemap, pscoast  
# Unix progs:   rm  
#  
ps=../example_01.ps  
gmtset GRID_CROSS_SIZE_PRIMARY 0 ANNOT_FONT_SIZE_PRIMARY 10  
psbasemap -R0/6.5/0/9 -Jx1i -B0 -P -K -U~Example 1 in Cookbook~ > $ps  
pscoast -Rg -JH0/6i -X0.25i -Y0.5i -O -K -Bg30 -Dc -Glightgray >> $ps  
grdcontour osu91a1f_16.nc -J -C10 -A50+s7 -Gd4i -L-1000/-1 -Wcthinnest,- -Wathin,- -O -K \  
        -T0.1i/0.02i >> $ps  
grdcontour osu91a1f_16.nc -J -C10 -A50+s7 -Gd4i -L-1/1000 -O -K -T0.1i/0.02i >> $ps  
pscoast -Rg -JH6i -Y4i -O -K -Bg30:.~Low Order Geoid~: -Dc -Glightgray >> $ps  
grdcontour osu91a1f_16.nc -J -C10 -A50+s7 -Gd4i -L-1000/-1 -Wcthinnest,- -Wathin,- -O -K \  
        -T0.1i/0.02i:-+ >> $ps  
grdcontour osu91a1f_16.nc -J -C10 -A50+s7 -Gd4i -L-1/1000 -O -T0.1i/0.02i:-+ >> $ps  
rm -f .gmt*
__________________________________________________________________________________________________

The first command draws a box surrounding the maps. This is followed by two sequences of pscoast, grdcontour, grdcontour. They differ in that the first is centered on Greenwich; the second on the dateline. We use the limit option (-L) in grdcontour to select negative contours only and plot those with a dashed pen, then positive contours only and draw with a solid pen [Default]. The -T option causes tickmarks pointing in the downhill direction to be drawn on the innermost, closed contours. For the upper panel we also added - and + to the local lows and highs. You can find this illustration as Figure 7.1.


PIC

Figure 7.1: Contour maps of gridded data.

7.2 Image presentations

As our second example we will demonstrate how to make color images from gridded data sets (again, we will defer the actual making of grid files to later examples). We will use the supplemental program grdraster to extract 2-D grid files of bathymetry and Geosat geoid heights and put the two images on the same page. The region of interest is the Hawaiian islands, and due to the oblique trend of the island chain we prefer to rotate our geographical data sets using an oblique Mercator projection defined by the hotspot pole at (68°W, 69°N). We choose the point (190°, 25.5°) to be the center of our projection (e.g., the local origin), and we want to image a rectangular region defined by the longitudes and latitudes of the lower left and upper right corner of region. In our case we choose (160°, 20°) and (220°, 30°) as the corners. We use grdimage to make the illustration:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 02  
#  
# Purpose:      Make two color images based gridded data  
# GMT progs:    gmtset, grd2cpt, grdgradient, grdimage, makecpt, psscale, pstext  
# Unix progs:   rm  
#  
ps=../example_02.ps  
gmtset HEADER_FONT_SIZE 30 OBLIQUE_ANNOTATION 0  
makecpt -Crainbow -T-2/14/2 > g.cpt  
grdimage HI_geoid2.nc -R160/20/220/30r -JOc190/25.5/292/69/4.5i -E50 -K -P \  
        -U/-1.25i/-1i/~Example 2 in Cookbook~ -B10 -Cg.cpt -X1.5i -Y1.25i > $ps  
psscale -Cg.cpt -D5.1i/1.35i/2.88i/0.4i -O -K -Ac -B2:GEOID:/:m: -E >> $ps  
grd2cpt HI_topo2.nc -Crelief -Z > t.cpt  
grdgradient HI_topo2.nc -A0 -Nt -GHI_topo2_int.nc  
grdimage HI_topo2.nc -IHI_topo2_int.nc -R -J -E50 -B10:.~H@#awaiian@# T@#opo and @#G@#eoid:~ -O -K \  
        -Ct.cpt -Y4.5i --HEADER_OFFSET=0.5i >> $ps  
psscale -Ct.cpt -D5.1i/1.35i/2.88i/0.4i -O -K -I0.3 -Ac -B2:TOPO:/:km: >> $ps  
pstext -R0/8.5/0/11 -Jx1i -O -N -Y-4.5i >> $ps << END  
-0.4 7.5 30 0.0 1 CB a)  
-0.4 3.0 30 0.0 1 CB b)  
END  
rm -f HI_topo2_int.nc ?.cpt .gmt*
__________________________________________________________________________________________________

The first step extracts the 2-D data sets from the local data base using grdraster, which is a supplemental utility program (see Appendix A) that may be adapted to reflect the nature of your data base format. It automatically figures out the required extent of the region given the two corners points and the projection. The extreme meridians and parallels enclosing the oblique region is -R159:50/220:10/3:10/47:35. This is the area extracted by grdraster. For your convenience we have commented out those lines and provided the two extracted files so you do not need grdraster to try this example. By using the embedded grid file format mechanism we saved the topography using kilometers as the data unit. We now have two grid files with bathymetry and geoid heights, respectively. We use makecpt to generate a linear color palette file geoid.cpt for the geoid and use grd2cpt to get a histogram-equalized cpt file topo.cpt for the topography data. To emphasize the structures in the data we calculate the slopes in the north-south direction using grdgradient; these will be used to modulate the color image. Next we run grdimage to create a color-code image of the Geosat geoid heights, and draw a color legend to the right of the image with psscale. Similarly, we run grdimage but specify -Y4.5i to plot above the previous image. Adding scale and label the two plots a) and b) completes the illustration (Figure 7.2).


PIC

Figure 7.2: Color images from gridded data.

7.3 Spectral estimation and xy-plots

In this example we will show how to use the GMT programs fitcircle, project, sample1d, spectrum1d, psxy, and pstext. Suppose you have (lon, lat, gravity) along a satellite track in a file called sat.xyg, and (lon, lat, gravity) along a ship track in a file called ship.xyg. You want to make a cross-spectral analysis of these data. First, you will have to get the two data sets into equidistantly sampled time-series form. To do this, it will be convenient to project these along the great circle that best fits the sat track. We must use fitcircle to find this great circle and choose the L2 estimates of best pole. We project the data using project to find out what their ranges are in the projected coordinate. The minmax utility will report the minimum and maximum values for multi-column ASCII tables. Use this information to select the range of the projected distance coordinate they have in common. The script prompts you for that information after reporting the values. We decide to make a file of equidistant sampling points spaced 1 km apart from -1167 to +1169, and use the UNIX utility $AWK to accomplish this step. We can then resample the projected data, and carry out the cross-spectral calculations, assuming that the ship is the input and the satellite is the output data. There are several intermediate steps that produce helpful plots showing the effect of the various processing steps (example_03[a–f].ps), while the final plot example_03.ps shows the ship and sat power in one diagram and the coherency on another diagram, both on the same page. Note the extended use of pstext and psxy to put labels and legends directly on the plots. For that purpose we often use -Jx1i and specify positions in inches directly. Thus, the complete automated script reads:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 03  
#  
# Purpose:      Resample track data, do spectral analysis, and plot  
# GMT progs:    filter1d, fitcircle, minmax, project, sample1d  
# GMT progs:    spectrum1d, trend1d, pshistogram, psxy, pstext  
# Unix progs:   $AWK, cat, echo, head, paste, rm, tail  
#  
# This example begins with data files ~ship.xyg~ and ~sat.xyg~ which  
# are measurements of a quantity ~g~ (a ~gravity anomaly~ which is an  
# anomalous increase or decrease in the magnitude of the acceleration  
# of gravity at sea level).  g is measured at a sequence of points ~x,y~  
# which in this case are ~longitude,latitude~.  The ~sat.xyg~ data were  
# obtained by a satellite and the sequence of points lies almost along  
# a great circle.  The ~ship.xyg~ data were obtained by a ship which  
# tried to follow the satellite’s path but deviated from it in places.  
# Thus the two data sets are not measured at the same of points,  
# and we use various GMT tools to facilitate their comparison.  
# The main illustration (example_03.ps) are accompanied with 5 support  
# plots (03a-f) showing data distributions and various intermediate steps.  
#  
# First, we use ~fitcircle~ to find the parameters of a great circle  
# most closely fitting the x,y points in ~sat.xyg~:  
#  
ps=../example_03.ps  
fitcircle sat.xyg -L2 > report  
cposx=‘grep ~L2 Average Position~ report | cut -f1‘  
cposy=‘grep ~L2 Average Position~ report | cut -f2‘  
pposx=‘grep ~L2 N Hemisphere~ report | cut -f1‘  
pposy=‘grep ~L2 N Hemisphere~ report | cut -f2‘  
#  
# Now we use ~project~ to project the data in both sat.xyg and ship.xyg  
# into data.pg, where g is the same and p is the oblique longitude around  
# the great circle.  We use -Q to get the p distance in kilometers, and -S  
# to sort the output into increasing p values.  
#  
project  sat.xyg -C$cposx/$cposy -T$pposx/$pposy -S -Fpz -Q > sat.pg  
project ship.xyg -C$cposx/$cposy -T$pposx/$pposy -S -Fpz -Q > ship.pg  
#  
# The minmax utility will report the minimum and maximum values for all columns.  
# We use this information first with a large -I value to find the appropriate -R  
# to use to plot the .pg data.  
#  
plotr=‘cat sat.pg ship.pg | minmax -I100/25‘  
psxy $plotr -U/-1.75i/-1.25i/~Example 3a in Cookbook~ \  
        -Ba500f100:~Distance along great circle~:/a100f25:~Gravity anomaly (mGal)~:WeSn \  
        -JX8i/5i -X2i -Y1.5i -K -Wthick sat.pg > ../example_03a.ps  
psxy -R -JX -O -Sp0.03i ship.pg >> ../example_03a.ps  
#  
# From this plot we see that the ship data have some ~spikes~ and also greatly  
# differ from the satellite data at a point about p ~= +250 km, where both of  
# them show a very large anomaly.  
#  
# To facilitate comparison of the two with a cross-spectral analysis using ~spectrum1d~,  
# we resample both data sets at intervals of 1 km.  First we find out how the data are  
# typically spaced using $AWK to get the delta-p between points and view it with  
# ~pshistogram~.  
#  
$AWK ’{ if (NR > 1) print $1 - last1; last1=$1; }’ ship.pg | pshistogram  -W0.1 -Gblack -JX3i -K \  
        -X2i -Y1.5i -B:.~Ship~: -U/-1.75i/-1.25i/~Example 3b in Cookbook~ > ../example_03b.ps  
$AWK ’{ if (NR > 1) print $1 - last1; last1=$1; }’ sat.pg  | pshistogram  -W0.1 -Gblack -JX3i -O \  
        -X5i -B:.~Sat~: >> ../example_03b.ps  
#  
# This experience shows that the satellite values are spaced fairly evenly, with  
# delta-p between 3.222 and 3.418.  The ship values are spaced quite unevelnly, with  
# delta-p between 0.095 and 9.017.  This means that when we want 1 km even sampling,  
# we can use ~sample1d~ to interpolate the sat data, but the same procedure applied  
# to the ship data could alias information at shorter wavelengths.  So we have to use  
# ~filter1d~ to resample the ship data.  Also, since we observed spikes in the ship  
# data, we use a median filter to clean up the ship values.  We will want to use ~paste~  
# to put the two sampled data sets together, so they must start and end at the same  
# point, without NaNs.  So we want to get a starting and ending point which works for  
# both of them.  This is a job for gmtmath UPPER/LOWER.  
#  
head -1 ship.pg > tmp  
head -1 sat.pg >> tmp  
sampr1=‘gmtmath tmp -Ca -Sf -F0 UPPER CEIL =‘  
tail -1 ship.pg > tmp  
tail -1 sat.pg >> tmp  
sampr2=‘gmtmath tmp -Ca -Sf -F0 LOWER FLOOR =‘  
#  
# Now we can use sampr1|2 in gmtmath to make a sampling points file for sample1d:  
gmtmath -T$sampr1/$sampr2/1 -N1/0 T = samp.x  
#  
# Now we can resample the projected satellite data:  
#  
sample1d sat.pg -Nsamp.x > samp_sat.pg  
#  
# For reasons above, we use filter1d to pre-treat the ship data.  We also need to sample it  
# because of the gaps > 1 km we found.  So we use filter1d | sample1d.  We also use the -E  
# on filter1d to use the data all the way out to sampr1/sampr2 :  
#  
filter1d ship.pg -Fm1 -T$sampr1/$sampr2/1 -E | sample1d -Nsamp.x > samp_ship.pg  
#  
# Now we plot them again to see if we have done the right thing:  
#  
psxy $plotr -JX8i/5i -X2i -Y1.5i -K -Wthick samp_sat.pg \  
        -Ba500f100:~Distance along great circle~:/a100f25:~Gravity anomaly (mGal)~:WeSn \  
        -U/-1.75i/-1.25i/~Example 3c in Cookbook~ > ../example_03c.ps  
psxy -R -JX -O -Sp0.03i samp_ship.pg >> ../example_03c.ps  
#  
# Now to do the cross-spectra, assuming that the ship is the input and the sat is the output  
# data, we do this:  
#  
paste samp_ship.pg samp_sat.pg | cut -f2,4 | spectrum1d -S256 -D1 -W -C > /dev/null  
#  
# Now we want to plot the spectra.  The following commands will plot the ship and sat  
# power in one diagram and the coherency on another diagram,  both on the same page.  
# Note the extended use of pstext and psxy to put labels and legends directly on the plots.  
# For that purpose we often use -Jx1i and specify positions in inches directly:  
#  
psxy spectrum.coh -Ba1f3p:~Wavelength (km)~:/a0.25f0.05:~Coherency@+2@+~:WeSn -JX-4il/3.75i \  
        -R1/1000/0/1 -U/-2.25i/-1.25i/~Example 3 in Cookbook~ -P -K -X2.5i -Sc0.07i -Gblack \  
        -Ey/0.5p -Y1.5i > $ps  
echo ~3.85 3.6 18 0.0 1 TR Coherency@+2@+~ | pstext -R0/4/0/3.75 -Jx1i -O -K >> $ps  
cat > box.d << END  
2.375   3.75  
2.375   3.25  
4       3.25  
END  
psxy -R -Jx -O -K -Wthicker box.d >> $ps  
psxy -Ba1f3p/a1f3p:~Power (mGal@+2@+km)~::.~Ship and Satellite Gravity~:WeSn spectrum.xpower \  
        -Gblack -ST0.07i -O -R1/1000/0.1/10000 -JX-4il/3.75il -Y4.2i -K -Ey/0.5p >> $ps  
psxy spectrum.ypower -R -JX -O -K -Gblack -Sc0.07i -Ey/0.5p >> $ps  
echo ~3.9 3.6 18 0.0 1 TR Input Power~ | pstext -R0/4/0/3.75 -Jx -O -K >> $ps  
psxy -R -Jx -O -K -Wthicker box.d >> $ps  
psxy -R -Jx -O -K -Glightgray -L -Wthicker >> $ps << END  
0.25    0.25  
1.4     0.25  
1.4     0.9  
0.25    0.9  
END  
echo ~0.4 0.7~ | psxy -R -Jx -O -K -ST0.07i -Gblack >> $ps  
echo ~0.5 0.7 14 0.0 1 ML Ship~ | pstext -R -Jx -O -K >> $ps  
echo ~0.4 0.4~ | psxy -R -Jx -O -K -Sc0.07i -Gblack >> $ps  
echo ~0.5 0.4 14 0.0 1 ML Satellite~ | pstext -R -Jx -O >> $ps  
#  
# Now we wonder if removing that large feature at 250 km would make any difference.  
# We could throw away a section of data with $AWK or sed or head and tail, but we  
# demonstrate the use of ~trend1d~ to identify outliers instead.  We will fit a  
# straight line to the samp_ship.pg data by an iteratively-reweighted method and  
# save the weights on output.  Then we will plot the weights and see how things  
# look:  
#  
trend1d -Fxw -N2r samp_ship.pg > samp_ship.xw  
psxy $plotr -JX8i/4i -X2i -Y1.5i -K -Sp0.03i \  
        -Ba500f100:~Distance along great circle~:/a100f25:~Gravity anomaly (mGal)~:WeSn \  
        -U/-1.75i/-1.25i/~Example 3d in Cookbook~ samp_ship.pg > ../example_03d.ps  
plotr=‘minmax samp_ship.xw -I100/1.1‘  
psxy $plotr -JX8i/1.1i -O -Y4.25i -Bf100/a0.5f0.1:~Weight~:Wesn -Sp0.03i samp_ship.xw \  
        >> ../example_03d.ps  
#  
# From this we see that we might want to throw away values where w < 0.6.  So we try that,  
# and this time we also use trend1d to return the residual from the model fit (the  
# de-trended data):  
trend1d -Fxrw -N2r samp_ship.pg | $AWK ’{ if ($3 > 0.6) print $1, $2 }’ \  
        | sample1d -Nsamp.x > samp2_ship.pg  
trend1d -Fxrw -N2r samp_sat.pg  | $AWK ’{ if ($3 > 0.6) print $1, $2 }’ \  
        | sample1d -Nsamp.x > samp2_sat.pg  
#  
# We plot these to see how they look:  
#  
plotr=‘cat samp2_sat.pg samp2_ship.pg | minmax -I100/25‘  
psxy $plotr -JX8i/5i -X2i -Y1.5i -K -Wthick \  
        -Ba500f100:~Distance along great circle~:/a50f25:~Gravity anomaly (mGal)~:WeSn \  
        -U/-1.75i/-1.25i/~Example 3e in Cookbook~ samp2_sat.pg > ../example_03e.ps  
psxy -R -JX -O -Sp0.03i samp2_ship.pg >> ../example_03e.ps  
#  
# Now we do the cross-spectral analysis again.  Comparing this plot (example_03e.ps) with  
# the previous one (example_03d.ps) we see that throwing out the large feature has reduced  
# the power in both data sets and reduced the coherency at wavelengths between 20--60 km.  
#  
paste samp2_ship.pg samp2_sat.pg | cut -f2,4 | spectrum1d -S256 -D1 -W -C > /dev/null  
#  
psxy spectrum.coh -Ba1f3p:~Wavelength (km)~:/a0.25f0.05:~Coherency@+2@+~:WeSn -JX-4il/3.75i \  
        -R1/1000/0/1 -U/-2.25i/-1.25i/~Example 3f in Cookbook~ -P -K -X2.5i -Sc0.07i -Gblack \  
        -Ey/0.5p -Y1.5i > ../example_03f.ps  
echo ~3.85 3.6 18 0.0 1 TR Coherency@+2@+~ | pstext -R0/4/0/3.75 -Jx -O -K >> ../example_03f.ps  
cat > box.d << END  
2.375   3.75  
2.375   3.25  
4       3.25  
END  
psxy -R -Jx -O -K -Wthicker box.d >> ../example_03f.ps  
psxy -Ba1f3p/a1f3p:~Power (mGal@+2@+km)~::.~Ship and Satellite Gravity~:WeSn spectrum.xpower \  
        -ST0.07i -O -R1/1000/0.1/10000 -JX-4il/3.75il -Y4.2i -K -Ey/0.5p >> ../example_03f.ps  
psxy spectrum.ypower -R -JX -O -K -Gblack -Sc0.07i -Ey/0.5p >> ../example_03f.ps  
echo ~3.9 3.6 18 0.0 1 TR Input Power~ | pstext -R0/4/0/3.75 -Jx -O -K >> ../example_03f.ps  
psxy -R -Jx -O -K -Wthicker box.d >> ../example_03f.ps  
psxy -R -Jx -O -K -Glightgray -L -Wthicker >> ../example_03f.ps << END  
0.25    0.25  
1.4     0.25  
1.4     0.9  
0.25    0.9  
END  
echo ~0.4 0.7~ | psxy -R -Jx -O -K -ST0.07i -Gblack >> ../example_03f.ps  
echo ~0.5 0.7 14 0.0 1 ML Ship~ | pstext -R -Jx -O -K >> ../example_03f.ps  
echo ~0.4 0.4~ | psxy -R -Jx -O -K -Sc0.07i -Gblack >> ../example_03f.ps  
echo ~0.5 0.4 14 0.0 1 ML Satellite~ | pstext -R -Jx -O >> ../example_03f.ps  
#  
rm -f box.d report tmp samp* *.pg *.extr spectrum.* .gmt*
__________________________________________________________________________________________________

The final illustration (Figure 7.3) shows that the ship gravity anomalies have more power than altimetry derived gravity for short wavelengths and that the coherency between the two signals improves dramatically for wavelengths > 20 km.


PIC

Figure 7.3: Spectral estimation and xy-plots.

7.4 A 3-D perspective mesh plot

This example will illustrate how to make a fairly complicated composite figure. We need a subset of the ETOPO5 bathymetry24 and Geosat geoid data sets which we will extract from the local data bases using grdraster. We would like to show a 2-layer perspective plot where layer one shows a contour map of the marine geoid with the location of the Hawaiian islands superposed, and a second layer showing the 3-D mesh plot of the topography. We also add an arrow pointing north and some text. This is how to do it:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 04  
#  
# Purpose:      3-D mesh plot of Hawaiian topography and geoid  
# GMT progs:    grdcontour, grdview, pscoast, pstext  
# Unix progs:   echo, rm  
#  
ps=../example_04.ps  
echo ’-10  255   0  255’ > zero.cpt  
echo ’  0  100  10  100’ >> zero.cpt  
grdcontour HI_geoid4.nc -Jm0.45i -E60/30 -R195/210/18/25 -C1 -A5+o -Gd4i -K -P -X1.5i -Y1.5i \  
        -U/-1.25i/-1.25i/~Example 4 in Cookbook~ > $ps  
pscoast -J -E60/30 -R -B2/2NEsw -Gblack -O -K -T209/19.5/1i >> $ps  
grdview HI_topo4.nc -J -Jz0.34i -Czero.cpt -E60/30 -R195/210/18/25/-6/4 -N-6/lightgray -Qsm -O -K \  
        -B2/2/2:~Topo (km)~:neswZ -Y2.2i >> $ps  
echo ’3.25 5.75 60 0.0 33 BC H@#awaiian@# R@#idge’ | pstext -R0/10/0/10 -Jx1i -O >> $ps  
rm -f zero.cpt .gmt*
__________________________________________________________________________________________________

The purpose of the color palette file zero.cpt is to have the positive topography mesh painted light gray (the remainder is white). The left side of Figure 7.4 shows the complete illustration.


PIC PIC

Figure 7.4: 3-D perspective mesh plot (left) and colored version (right).

A color version of this figure was used in our first article in EOS Trans. AGU (Oct. 8th, 1991). It was created along similar lines, but instead of a mesh plot we chose a color-coded surface with artificial illumination from a light-source due north. We choose to use the -Qi option in grdview to achieve a high degree of smoothness. Here, we select 100 dpi since that will be the resolution of our final raster (The EOS raster was 300 dpi). We used grdgradient to provide the intensity files. The following script creates the color PostScript file. Note that the size of the resulting output file is directly dependent on the square of the dpi chosen for the scanline conversion. A higher value for dpi in -Qi would have resulted in a much larger output file. The cpt files were taken from Section 7.2.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 04c  
#  
# 3-D perspective color plot of Hawaiian topography and geoid  
# GMT progs:    grdcontour, grdview, pscoast, pstext  
# Unix progs:   echo, rm  
#  
ps=../example_04c.ps  
grdgradient HI_geoid4.nc -A0 -Gg_intens.nc -Nt0.75 -M  
grdgradient HI_topo4.nc -A0 -Gt_intens.nc -Nt0.75 -M  
grdview HI_geoid4.nc -Ig_intens.nc -JM6.75i -E60/30 -R195/210/18/25 -Cgeoid.cpt -Qi100 -K \  
        -X1.5i -Y1.25i -P -U/-1.25i/-1i/~Example 04c in Cookbook~ > $ps  
pscoast -J -E60/30 -R -B2/2NEsw -Gblack -O -K >> $ps  
psbasemap -R -J -E60/30 -O -K -T209/19.5/1i --COLOR_BACKGROUND=red --TICK_PEN=thinner,red >> $ps  
grdview HI_topo4.nc -It_intens.nc -J -JZ3.4i -Ctopo.cpt -E60/30 -R195/210/18/25/-6/4 \  
        -N-6/lightgray -Qi100 -O -K -Y2.2i >> $ps  
psbasemap -J -JZ3.4i -E60/30 -R -Z-6 -O -K -B2/2/2:~Topo (km)~:neZ >> $ps  
echo ’3.25 5.75 60 0.0 33 BC H@#awaiian@# R@#idge’ | pstext -R0/10/0/10 -Jx1i -O >> $ps  
rm -f *_intens.nc .gmt*
__________________________________________________________________________________________________

7.5 A 3-D illuminated surface in black and white

Instead of a mesh plot we may choose to show 3-D surfaces using artificial illumination. For this example we will use grdmath to make a grid file that contains the surface given by the function z(x,y) = cos(2πr8)e-r10, where r2 = (x2 +y2). The illumination is obtained by passing two grid files to grdview: One with the z-values (the surface) and another with intensity values (which should be in the ±1 range). We use grdgradient to compute the horizontal gradients in the direction of the artificial light source. The gray.cpt file only has one line that states that all z values should have the gray level 128. Thus, variations in shade are entirely due to variations in gradients, or illuminations. We choose to illuminate from the SW and view the surface from SE:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 05  
#  
# Purpose:      Generate grid and show monochrome 3-D perspective  
# GMT progs:    grdgradient, grdmath, grdview, pstext  
# Unix progs:   echo, rm  
#  
ps=../example_05.ps  
grdmath -R-15/15/-15/15 -I0.3 X Y HYPOT DUP 2 MUL PI MUL 8 DIV COS EXCH NEG 10 DIV EXP MUL = \  
        sombrero.nc  
echo ’-5 128 5 128’ > gray.cpt  
grdgradient sombrero.nc -A225 -Gintensity.nc -Nt0.75  
grdview sombrero.nc -JX6i -JZ2i -B5/5/0.5SEwnZ -N-1/white -Qs -Iintensity.nc -X1.5i -Cgray.cpt \  
        -R-15/15/-15/15/-1/1 -K -E120/30 -U/-1.25i/-0.75i/~Example 5 in Cookbook~ > $ps  
echo ~4.1 5.5 50 0 33 BC z(r) = cos (2@~p@~r/8) * e@+-r/10@+~ | pstext -R0/11/0/8.5 -Jx1i -O >> $ps  
rm -f gray.cpt sombrero.nc intensity.nc .gmt*
__________________________________________________________________________________________________

The variations in intensity could be made more dramatic by using grdmath to scale the intensity file before running grdview. For very rough data sets one may improve the smoothness of the intensities by passing the output of grdgradient to grdhisteq. The shell-script above will result in a plot like the one in Figure 7.5.


PIC

Figure 7.5: 3-D illuminated surface.

7.6 Plotting of histograms

GMT provides two tools to render histograms: pshistogram and psrose. The former takes care of regular histograms whereas the latter deals with polar histograms (rose diagrams, sector diagrams, and wind rose diagrams). We will show an example that involves both programs. The file fractures.yx contains a compilation of fracture lengths and directions as digitized from geological maps. The file v3206.t contains all the bathymetry measurements from Vema cruise 3206. Our complete figure (Figure 7.6) was made running this script:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 06  
#  
# Purpose:      Make standard and polar histograms  
# GMT progs:    pshistogram, psrose  
# Unix progs:   rm  
#  
ps=../example_06.ps  
psrose fractures.d -: -A10r -S1.8in -U/-2.25i/-0.75i/~Example 6 in Cookbook~ -P -Gblack -R0/1/0/360 \  
        -X2.5i -K -B0.2g0.2/30g30 > $ps  
pshistogram -Ba2000f1000:~Topography (m)~:/a10f5:~Frequency~::,%::.~Two types of histograms~:WSne \  
        v3206.t -R-6000/0/0/30 -JX4.8i/2.4i -Ggray -O -Y5.5i -X-0.5i -Lthinner -Z1 -W250 >> $ps  
rm -f .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.6: Two kinds of histograms.

7.7 A simple location map

Many scientific papers start out by showing a location map of the region of interest. This map will typically also contain certain features and labels. This example will present a location map for the equatorial Atlantic ocean, where fracture zones and mid-ocean ridge segments have been plotted. We also would like to plot earthquake locations and available isochrons. We have obtained one file, quakes.xym, which contains the position and magnitude of available earthquakes in the region. We choose to use magnitude/100 for the symbol-size in inches. The digital fracture zone traces (fz.xy) and isochrons (0 isochron as ridge.xy, the rest as isochrons.xy) were digitized from available maps25. We create the final location map (Figure 7.7) with the following script:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 07  
#  
# Purpose:      Make a basemap with earthquakes and isochrons etc  
# GMT progs:    pscoast, pstext, psxy  
# Unix progs:   $AWK, echo, rm  
#  
ps=../example_07.ps  
pscoast -R-50/0/-10/20 -JM9i -K -GP300/26 -Dl -Wthinnest -B10 -U~Example 7 in Cookbook~ > $ps  
psxy -R -J -O -K -m fz.xy -Wthinner,- >> $ps  
$AWK ’{print $1-360.0, $2, $3*0.01}’ quakes.xym | psxy -R -J -O -K -H1 -Sci -Gwhite -Wthinnest >> $ps  
psxy -R -J -O -K -m isochron.xy -Wthin >> $ps  
psxy -R -J -O -K -m ridge.xy -Wthicker >> $ps  
psxy -R -J -O -K -Gwhite -Wthick -A >> $ps << END  
-14.5   15.2  
 -2     15.2  
 -2     17.8  
-14.5   17.8  
END  
psxy -R -J -O -K -Gwhite -Wthinner -A >> $ps << END  
-14.35  15.35  
 -2.15  15.35  
 -2.15  17.65  
-14.35  17.65  
END  
echo ~-13.5 16.5~ | psxy -R -J -O -K -Sc0.08i -Gwhite -Wthinner >> $ps  
echo ~-12.5 16.5 18 0 6 LM ISC Earthquakes~ | pstext -R -J -O -K >> $ps  
pstext -R -J -O -Sthin -Gwhite >> $ps << END  
-43 -5 30 0 1 CM SOUTH  
-43 -8 30 0 1 CM AMERICA  
 -7 11 30 0 1 CM AFRICA  
END  
rm -f .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.7: A typical location map.

The same figure could equally well be made in color, which could be rasterized and made into a slide for a meeting presentation. The script is similar to the one outlined above, except we would choose a color for land and oceans, and select colored symbols and pens rather than black and white.

7.8 A 3-D histogram

The program psxyz allows us to plot three-dimensional symbols, including columnar plots. As a simple demonstration, we will convert a gridded netCDF of bathymetry into an ASCII xyz table and use the height information to draw a 2-D histogram in a 3-D perspective view. Our gridded bathymetry file is called guinea_bay.nc and covers the region from 0 to 5 °E and 0 to 5 °N. Depth ranges from -5000 meter to sea-level. We produce the Figure 7.8 by running this script:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 08  
#  
# Purpose:      Make a 3-D bar plot  
# GMT progs:    grd2xyz, pstext, psxyz  
# Unix progs:   echo, rm  
#  
ps=../example_08.ps  
grd2xyz guinea_bay.nc | psxyz -B1/1/1000:~Topography (m)~::.ETOPO5:WSneZ+ \  
        -R-0.1/5.1/-0.1/5.1/-5000/0 -JM5i -JZ6i -E200/30 -So0.0833333ub-5000 -P \  
        -U~Example 8 in Cookbook~ -Wthinnest -Glightgray -K > $ps  
echo ’0.1 4.9 24 0 1 TL This is the surface of cube’ | pstext -R -J -JZ -Z0 -E200/30 -O >> $ps  
rm -f .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.8: A 3-D histogram.

7.9 Plotting time-series along tracks

A common application in many scientific disciplines involves plotting one or several time-series as as “wiggles” along tracks. Marine geophysicists often display magnetic anomalies in this manner, and seismologists use the technique when plotting individual seismic traces. In our example we will show how a set of Geosat sea surface slope profiles from the south Pacific can be plotted as “wiggles” using the pswiggle program. We will embellish the plot with track numbers, the location of the Pacific-Antarctic Ridge, recognized fracture zones in the area, and a “wiggle” scale. The Geosat tracks are stored in the files *.xys, the ridge in ridge.xy, and all the fracture zones are stored in the multiple segment file fz.xy. We extract the profile id (which is the first part of the file name for each profile) and the last point in each of the track files to construct an input file for pstext that will label each profile with the track number. We know the profiles trend approximately N40°E so we want the labels to have that same orientation (i.e., the angle with the baseline must be 50°). We do this by extracting the last record from each track, paste this file with the tracks.lis file, and use $AWK to create the format needed for pstext. Note we offset the positions by -0.05 inch with -D in order to have a small gap between the profile and the label:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 09  
#  
# Purpose:      Make wiggle plot along track from geoid deflections  
# GMT progs:    pswiggle, pstext, psxy  
# Unix progs:   $AWK, ls, paste, tail, rm  
#  
ps=../example_09.ps  
pswiggle track_*.xys -R185/250/-68/-42 -U~Example 9 in Cookbook~ -K -Jm0.13i -Ba10f5 -Gblack \  
        -Z2000 -Wthinnest -S240/-67/500/@~m@~rad > $ps  
psxy -R -J -O -K ridge.xy -Wthicker >> $ps  
psxy -R -J -O -K -m fz.xy -Wthinner,- >> $ps  
rm -f tmp  
# Make label file  
for file in track_*.xys; do  
        tail -1 $file >> tmp  
done  
ls -1 track_*.xys | $AWK -F. ’{print $2}’ > tracks.lis  
paste tmp tracks.lis | $AWK ’{print $1, $2, 10, 50, 1, ~RM~, $4}’ \  
        | pstext -R -J -D-0.05i/-0.05i -O >> $ps  
rm -f tmp tracks.lis .gmt*
__________________________________________________________________________________________________

The output shows the sea-surface slopes along 42 descending Geosat tracks in the Eltanin and Udintsev fracture zone region in a Mercator projection (Figure 7.9).


PIC

Figure 7.9: Time-series as “wiggles” along a track.

7.10 A geographical bar graph plot

Our next and perhaps most business-like example presents a three-dimensional bar graph plot showing the geographic distribution of the membership in the American Geophysical Union (AGU). The input data was taken from the January 2008 AGU member directory and added up to give total members per continent. We decide to plot a 3-D column centered on each continent with a height that is proportional to the logarithm of the membership. A log10-scale is used since the memberships vary by almost 3 orders of magnitude. We choose a plain linear projection for the basemap and add the columns and text on top. Our script that produces Figure 7.10 reads:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 10  
#  
# Purpose:      Make 3-D bar graph on top of perspective map  
# GMT progs:    pscoast, pstext, psxyz  
# Unix progs:   $AWK, rm  
#  
ps=../example_10.ps  
pscoast -Rd -JX8id/5id -Dc -Gblack -E200/40 -K -U~Example 10 in Cookbook~ > $ps  
psxyz agu2013.txt -R-180/180/-90/90/1/100000 -J -JZ2.5il -So0.3ib1 -Ggray -Wthinner \  
        -B60g60/30g30/a1p:Memberships:WSneZ -O -K -E200/40 >> $ps  
$AWK ’{print $1, $2, 20, 0, 0, ~RM~, $3}’ agu2013.txt \  
        | pstext -Rd -J -O -K -E200/40 -Gwhite -Sthinner -D-0.2i/0 >> $ps  
echo ~4.5 6 30 0 5 BC AGU 2013 Membership Distribution~ | pstext -R0/11/0/8.5 -Jx1i -O >> $ps  
rm -f .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.10: Geographical bar graph.

7.11 Making a 3-D RGB color cube

In this example we generate a series of 6 color images, arranged so that they can be cut out and assembled into a 3-D color cube. The six faces of the cube represent the outside of the R-G-B color space. On each face one of the color components is fixed at either 0 or 255 and the other two components vary smoothly across the face from 0 to 255. The cube is configured as a right-handed coordinate system with x-y-z mapping R-G-B. Hence, the 8 corners of the cube represent the primaries red, green, and blue, plus the secondaries cyan, magenta and yellow, plus black and white.

The 6 color faces are generated by feeding grdimage three grids, one for each color component (R, G, and B). In some cases the X or Y axes of a face are reversed by specifying a negative width or height in order to change the variation of the color value in that direction from ascending to descending, or vice versa.

A number of rays emanating from the white and black corners indicate the Hue value (ranging from 0 to 360°). The dashed and dotted lines near the white corner reflect saturation levels, running from 0 to 1 (in black font). On these 3 faces the brightness is a constant value of 1. On the other 3 faces of the cube, around the black corner, the white decimal numbers indicate brightnesses between 0 and 1, with saturation fixed at 1.

Here is the shell script to generate the RGB cube in Figure 7.11:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 11  
#  
# Purpose:      Create a 3-D RGB Cube  
# GMT progs:    gmtset, grdimage, grdmath, pstext, psxy  
# Unix progs:   rm  
ps=../example_11.ps  
 
# Use psxy to plot ~cut-along-the-dotted~ lines.  
 
gmtset TICK_LENGTH 0 COLOR_MODEL rgb CHAR_ENCODING Standard+  
 
psxy cut-here.dat -Wthinnest,. -m -R-51/306/0/1071 -JX3.5i/10.5i -X2.5i -Y0.5i \  
        -P -U/-2.0i/-0.2i/~Example 11 in Cookbook~ -K > $ps  
 
# First, create grids of ascending X and Y and constant 0.  
# These are to be used to represent R, G and B values of the darker 3 faces of the cube.  
 
grdmath -I1 -R0/255/0/255 X = x.nc  
grdmath -I1 -R Y = y.nc  
grdmath -I1 -R 0 = c.nc  
 
grdimage x.nc y.nc c.nc -JX2.5i/-2.5i -R -K -O -X0.5i >> $ps  
psxy -m -Wthinner,white,- rays.dat -J -R -K -O -Bwesn >> $ps  
echo 128 128 12 -45 1 MC ~60\217~ | pstext -Gwhite -J -R -K -O >> $ps  
echo 102 26 12 -90 1 MC 0.4 | pstext -Gwhite -J -R -K -O >> $ps  
echo 204 26 12 -90 1 MC 0.8 | pstext -Gwhite -J -R -K -O >> $ps  
echo 10 140 16 180 1 MC G | pstext -Gwhite -J -R -K -O >> $ps  
echo 0 0 0 128 | psxy -N -Svs -Gwhite -J -R -K -O >> $ps  
 
grdimage x.nc c.nc y.nc -JX2.5i/2.5i -R -K -O -Y2.5i >> $ps  
psxy -m -Wthinner,white,- rays.dat -J -R -K -O -Bwesn >> $ps  
echo 128 128 12 45 1 MC ~300\217~ | pstext -Gwhite -J -R -K -O >> $ps  
echo 26 102 12 0 1 MC 0.4 | pstext -Gwhite -J -R -K -O >> $ps  
echo 26 204 12 0 1 MC 0.8 | pstext -Gwhite -J -R -K -O >> $ps  
echo 140 10 16 -90 1 MC R | pstext -Gwhite -J -R -K -O >> $ps  
echo 100 100 16 -45 1 MC V | pstext -Gwhite -J -R -K -O >> $ps  
echo 0 0 128 0 | psxy -N -Svs -Gwhite -J -R -K -O >> $ps  
echo 0 0 90 90 | psxy -N -Svs -Gwhite -J -R -K -O >> $ps  
 
grdimage c.nc x.nc y.nc -JX-2.5i/2.5i -R -K -O -X-2.5i >> $ps  
psxy -m -Wthinner,white,- rays.dat -J -R -K -O -Bwesn >> $ps  
echo 128 128 12 135 1 MC ~180\217~ | pstext -Gwhite -J -R -K -O >> $ps  
echo 102 26 12 90 1 MC 0.4 | pstext -Gwhite -J -R -K -O >> $ps  
echo 204 26 12 90 1 MC 0.8 | pstext -Gwhite -J -R -K -O >> $ps  
echo 10 140 16 0 1 MC B | pstext -Gwhite -J -R -K -O >> $ps  
echo 0 0 0 128 | psxy -N -Svs -Gwhite -J -R -K -O >> $ps  
echo 0 0 128 0 | psxy -N -Svs -Gwhite -J -R -K -O >> $ps  
 
# Second, create grids of descending X and Y and constant 255.  
# These are to be used to represent R, G and B values of the lighter 3 faces of the cube.  
 
grdmath -I1 -R 255 X SUB = x.nc  
grdmath -I1 -R 255 Y SUB = y.nc  
grdmath -I1 -R 255       = c.nc  
 
grdimage x.nc y.nc c.nc -JX-2.5i/-2.5i -R -K -O -X2.5i -Y2.5i >> $ps  
psxy -m -Wthinner,black,- rays.dat -J -R -K -O -Bwesn >> $ps  
echo 128 128 12 225 1 MC ~240\217~ | pstext -J -R -K -O >> $ps  
echo 102 26 12 270 1 MC 0.4 | pstext -J -R -K -O >> $ps  
echo 204 26 12 270 1 MC 0.8 | pstext -J -R -K -O >> $ps  
 
grdimage c.nc y.nc x.nc -JX2.5i/-2.5i -R -K -O -X2.5i >> $ps  
psxy -m -Wthinner,black,- rays.dat -J -R -K -O -Bwesn >> $ps  
echo 128 128 12 -45 1 MC ~0\217~ | pstext -J -R -K -O >> $ps  
echo 26 102 12 0 1 MC 0.4 | pstext -J -R -K -O >> $ps  
echo 26 204 12 0 1 MC 0.8 | pstext -J -R -K -O >> $ps  
echo 100 100 16 45 1 MC S | pstext -Gblack -J -R -K -O >> $ps  
echo 204 66 16 90 1 MC H | pstext -Gblack -J -R -K -O >> $ps  
echo 0 0 90 90 | psxy -N -Svs -Gblack -J -R -K -O >> $ps  
echo 204 204 204 76 | psxy -N -Svs -Gblack -J -R -K -O >> $ps  
 
grdimage x.nc c.nc y.nc -JX-2.5i/2.5i -R -K -O -X-2.5i -Y2.5i >> $ps  
psxy -m -Wthinner,black,- rays.dat -J -R -K -O -Bwesn >> $ps  
echo 128 128 12 135 1 MC ~120\217~ | pstext -J -R -K -O >> $ps  
echo 26 102 12 180 1 MC 0.4 | pstext -J -R -K -O >> $ps  
echo 26 204 12 180 1 MC 0.8 | pstext -J -R -K -O >> $ps  
echo 200 200 16 225 1 MC GMT 4 | pstext -J -R -O >> $ps  
 
rm -f *.nc .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.11: The RGB color cube.

7.12 Optimal triangulation of data

Our next example (Figure 7.12) operates on a data set of topographic readings non-uniformly distributed in the plane (Table 5.11 in Davis: Statistics and Data Analysis in Geology, J. Wiley). We use triangulate to perform the optimal Delaunay triangulation, then use the output to draw the resulting network. We label the node numbers as well as the node values, and call pscontour to make a contour map and image directly from the raw data. Thus, in this example we do not actually make grid files but still are able to contour and image the data. We use a color palette table topo.cpt (created via minmax and makecpt). The script becomes:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 12  
#  
# Purpose:      Illustrates Delaunay triangulation of points, and contouring  
# GMT progs:    makecpt, minmax, pscontour, pstext, psxy, triangulate  
# Unix progs:   $AWK, echo, rm  
#  
# First draw network and label the nodes  
#  
ps=../example_12.ps  
triangulate table_5.11 -m > net.xy  
psxy -R0/6.5/-0.2/6.5 -JX3.06i/3.15i -B2f1WSNe -m net.xy -Wthinner -P -K -X0.9i -Y4.65i > $ps  
psxy table_5.11 -R -J -O -K -Sc0.12i -Gwhite -Wthinnest >> $ps  
$AWK ’{print $1, $2, 6, 0, 0, ~CM~, NR-1}’ table_5.11 | pstext -R -J -O -K >> $ps  
#  
# Then draw network and print the node values  
#  
psxy -R -J -B2f1eSNw -m net.xy -Wthinner -O -K -X3.25i >> $ps  
psxy -R -J -O -K table_5.11 -Sc0.03i -Gblack >> $ps  
$AWK ’{printf ~%g %s 6 0 0 LM %g\n~, $1, $2, $3}’ table_5.11 | pstext -R -J -O -K -Wwhite,o \  
        -C0.01i/0.01i -D0.08i/0i -N >> $ps  
#  
# Then contour the data and draw triangles using dashed pen; use ~minmax~ and ~makecpt~ to make a  
# color palette (.cpt) file  
#  
T=‘minmax -T25/2 table_5.11‘  
makecpt -Cjet $T > topo.cpt  
pscontour -R -J table_5.11 -B2f1WSne -Wthin -Ctopo.cpt -Lthinnest,- -G1i/0 -X-3.25i -Y-3.65i -O -K \  
        -U~Example 12 in Cookbook~ >> $ps  
#  
# Finally color the topography  
#  
pscontour -R -J table_5.11 -B2f1eSnw -Ctopo.cpt -I -X3.25i -O -K >> $ps  
echo ~3.16 8 30 0 1 BC Delaunay Triangulation~ | \  
        pstext -R0/8/0/11 -Jx1i -O -X-3.25i >> $ps  
#  
rm -f net.xy topo.cpt .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.12: Optimal triangulation of data.

7.13 Plotting of vector fields

In many areas, such as fluid dynamics and elasticity, it is desirable to plot vector fields of various kinds. GMT provides a way to illustrate 2-component vector fields using the grdvector utility. The two components of the field (Cartesian or polar components) are stored in separate grid files. In this example we use grdmath to generate a surface z(x,y) = xexp(-x2 -y2) and to calculate z by returning the x- and y-derivatives separately. We superpose the gradient vector field and the surface z and also plot the components of the gradient in separate windows. A pstext call to place a header finishes the plot (Figure 7.13:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 13  
#  
# Purpose:      Illustrate vectors and contouring  
# GMT progs:    grdmath, grdcontour, grdvector, pstext  
# Unix progs:   echo, rm  
#  
ps=../example_13.ps  
grdmath -R-2/2/-2/2 -I0.1 X Y R2 NEG EXP X MUL = z.nc  
grdmath z.nc DDX = dzdx.nc  
grdmath z.nc DDY = dzdy.nc  
grdcontour dzdx.nc -JX3i -B1/1WSne -C0.1 -A0.5 -K -P -Gd2i -S4 -T0.1i/0.03i \  
        -U~Example 13 in Cookbook~ > $ps  
grdcontour dzdy.nc -J -B1/1WSne -C0.05 -A0.2 -O -K -Gd2i -S4 -T0.1i/0.03i -X3.45i >> $ps  
grdcontour z.nc -J -B1/1WSne -C0.05 -A0.1 -O -K -Gd2i -S4 -T0.1i/0.03i -X-3.45i -Y3.45i >> $ps  
grdcontour z.nc -J -B1/1WSne -C0.05 -O -K -Gd2i -S4 -X3.45i >> $ps  
grdvector dzdx.nc dzdy.nc -I0.2 -J -O -K -Q0.03i/0.1i/0.09in0.25i -G0 -S5i >> $ps  
echo ~3.2 3.6 40 0 6 BC z(x,y) = x * exp(-x@+2@+-y@+2@+)~ \  
        | pstext -R0/6/0/4.5 -Jx1i -O -X-3.45i >> $ps  
rm -f z.nc dzdx.nc dzdy.nc .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.13: Display of vector fields in GMT.

7.14 Gridding of data and trend surfaces

This example shows how one goes from randomly spaced data points to an evenly sampled surface. First we plot the distribution and values of our raw data set (same as in Section 7.12). We choose an equidistant grid and run blockmean which preprocesses the data to avoid aliasing. The dashed lines indicate the logical blocks used by blockmean; all points inside a given bin will be averaged. The logical blocks are drawn from a temporary file we make on the fly within the shell script. The processed data is then gridded with the surface program and contoured every 25 units. A most important point here is that blockmean, blockmedian, or blockmode should always be run prior to running surface, and both of these steps must use the same grid interval. We use grdtrend to fit a bicubic trend surface to the gridded data, contour it as well, and sample both grid files along a diagonal transect using grdtrack. The bottom panel compares the gridded (solid line) and bicubic trend (dashed line) along the transect using psxy (Figure 7.14):

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 14  
#  
# Purpose:      Showing simple gridding, contouring, and resampling along tracks  
# GMT progs:    blockmean, grdcontour, grdtrack, grdtrend, minmax, project  
# GMT progs:    gmtset, pstext, psbasemap, psxy, surface  
# Unix progs:   $AWK, rm  
#  
ps=../example_14.ps  
 
# First draw network and label the nodes  
 
gmtset GRID_PEN_PRIMARY thinnest,-  
psxy table_5.11 -R0/7/0/7 -JX3.06i/3.15i -B2f1WSNe -Sc0.05i -Gblack -P -K -Y6.45i > $ps  
$AWK ’{printf ~%g %s 6 0 0 LM %g\n~, $1+0.08, $2, $3}’ table_5.11 | pstext -R -J -O -K -N >> $ps  
blockmean table_5.11 -R0/7/0/7 -I1 > mean.xyz  
 
# Then draw blockmean cells  
 
psbasemap -R0.5/7.5/0.5/7.5 -J -O -K -B0g1 -X3.25i >> $ps  
psxy -R0/7/0/7 -J -B2f1eSNw mean.xyz -Ss0.05i -Gblack -O -K >> $ps  
$AWK ’{printf ~%g %s 6 0 0 LM %g\n~, $1+0.1, $2, $3}’ mean.xyz \  
        | pstext -R -J -O -K -Wwhite,o -C0.01i/0.01i -N >> $ps  
 
# Then surface and contour the data  
 
surface mean.xyz -R -I1 -Gdata.nc  
grdcontour data.nc -J -B2f1WSne -C25 -A50 -G3i/10 -S4 -O -K -X-3.25i -Y-3.55i >> $ps  
psxy -R -J mean.xyz -Ss0.05i -Gblack -O -K >> $ps  
 
# Fit bicubic trend to data and compare to gridded surface  
 
grdtrend data.nc -N10 -Ttrend.nc  
project -C0/0 -E7/7 -G0.1 -N > track  
grdcontour trend.nc -J -B2f1wSne -C25 -A50 -Glct/cb -S4 -O -K -X3.25i >> $ps  
psxy -R -J track -Wthick,. -O -K >> $ps  
 
# Sample along diagonal  
 
grdtrack track -Gdata.nc | cut -f3,4 > data.d  
grdtrack track -Gtrend.nc | cut -f3,4 > trend.d  
psxy ‘minmax data.d trend.d -I0.5/25‘ -JX6.3i/1.4i data.d -Wthick -O -K -X-3.25i -Y-1.9i \  
        -B1/50WSne >> $ps  
psxy -R -J trend.d -Wthinner,- -O -U~Example 14 in Cookbook~ >> $ps  
 
rm -f mean.xyz track *.nc *.d .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.14: Gridding of data and trend surfaces.

7.15 Gridding, contouring, and masking of unconstrained areas

This example (Figure 7.15) demonstrates some off the different ways one can use to grid data in GMT, and how to deal with unconstrained areas. We first convert a large ASCII file to binary with gmtconvert since the binary file will read and process much faster. Our lower left plot illustrates the results of gridding using a nearest neighbor technique (nearneighbor) which is a local method: No output is given where there are no data. Next (lower right), we use a minimum curvature technique (surface) which is a global method. Hence, the contours cover the entire map although the data are only available for portions of the area (indicated by the gray areas plotted using psmask). The top left scenario illustrates how we can create a clip path (using psmask) based on the data coverage to eliminate contours outside the constrained area. Finally (top right) we simply employ pscoast to overlay gray land masses to cover up the unwanted contours, and end by plotting a star at the deepest point on the map with psxy. This point was extracted from the grid files using grdinfo.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 15  
#  
# Purpose:      Gridding and clipping when data are missing  
# GMT progs:    blockmedian, gmtconvert, grdclip, grdcontour, grdinfo, minmax  
# GMT progs:    nearneighbor, pscoast, psmask, pstext, surface  
# Unix progs:   awk, echo, rm  
#  
ps=../example_15.ps  
gmtconvert ship.xyz -bo > ship.b  
#  
region=‘minmax ship.b -I1 -bi3‘  
nearneighbor $region -I10m -S40k -Gship.nc ship.b -bi3  
info=‘grdinfo -C -M ship.nc‘  
grdcontour ship.nc -JM3i -P -B2WSne -C250 -A1000 -G2i -K -U~Example 15 in Cookbook~ > $ps  
#  
blockmedian $region -I10m ship.b -bi3 -bo > ship_10m.b  
surface $region -I10m ship_10m.b -Gship.nc -bi3  
psmask $region -I10m ship.b -J -O -K -T -Glightgray -bi3 -X3.6i >> $ps  
grdcontour ship.nc -J -B2WSne -C250 -L-8000/0 -A1000 -G2i -O -K >> $ps  
#  
psmask $region -I10m ship_10m.b -bi3 -J -B2WSne -O -K -X-3.6i -Y3.75i >> $ps  
grdcontour ship.nc -J -C250 -A1000 -L-8000/0 -G2i -O -K >> $ps  
psmask -C -O -K >> $ps  
#  
grdclip ship.nc -Sa-1/NaN -Gship_clipped.nc  
grdcontour ship_clipped.nc -J -B2WSne -C250 -A1000 -L-8000/0 -G2i -O -K -X3.6i >> $ps  
pscoast $region -J -O -K -Ggray -Wthinnest >> $ps  
echo $info | $AWK ’{print $12,$13}’ | psxy -R -J -O -K -Sa0.15i -Wthick >> $ps  
echo ~-0.3 3.6 24 0 1 CB Gridding with missing data~ | pstext -R0/3/0/4 -Jx1i -O -N >> $ps  
rm -f ship.b ship_10m.b ship.nc ship_clipped.nc .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.15: Gridding, contouring, and masking of data.

7.16 Gridding of data, continued

pscontour (for contouring) and triangulate (for gridding) use the simplest method of interpolating data: a Delaunay triangulation (see Section 7.12) which forms z(x,y) as a union of planar triangular facets. One advantage of this method is that it will not extrapolate z(x,y) beyond the convex hull of the input (x, y) data. Another is that it will not estimate a z value above or below the local bounds on any triangle. A disadvantage is that the z(x,y) surface is not differentiable, but has sharp kinks at triangle edges and thus also along contours. This may not look physically reasonable, but it can be filtered later (last panel below). surface can be used to generate a higher-order (smooth and differentiable) interpolation of z(x,y) onto a grid, after which the grid may be illustrated (grdcontour, grdimage, grdview). surface will interpolate to all (x, y) points in a rectangular region, and thus will extrapolate beyond the convex hull of the data. However, this can be masked out in various ways (see Section 7.15).

A more serious objection is that surface may estimate z values outside the local range of the data (note area near x = 0.8, y = 5.3). This commonly happens when the default tension value of zero is used to create a “minimum curvature” (most smooth) interpolant. surface can be used with non-zero tension to partially overcome this problem. The limiting value tension = 1 should approximate the triangulation, while a value between 0 and 1 may yield a good compromise between the above two cases. A value of 0.5 is shown here (Figure 7.16). A side effect of the tension is that it tends to make the contours turn near the edges of the domain so that they approach the edge from a perpendicular direction. A solution is to use surface in a larger area and then use grdcut to cut out the desired smaller area. Another way to achieve a compromise is to interpolate the data to a grid and then filter the grid using grdfft or grdfilter. The latter can handle grids containing “NaN” values and it can do median and mode filters as well as convolutions. Shown here is triangulate followed by grdfilter. Note that the filter has done some extrapolation beyond the convex hull of the original x, y values. The “best” smooth approximation of z(x,y) depends on the errors in the data and the physical laws obeyed by z. GMT cannot always do the “best” thing but it offers great flexibility through its combinations of tools. We illustrate all four solutions using a cpt file that contains color fills, predefined patterns for interval (900,925) and NaN, an image pattern for interval (875,900), and a “skip slice” request for interval (700,725).

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 16  
#  
# Purpose:      Illustrates interpolation methods using same data as Example 12.  
# GMT progs:    gmtset, grdview, grdfilter, pscontour, psscale, pstext, surface, triangulate  
# Unix progs:   echo, rm  
#  
# Illustrate various means of contouring, using triangulate and surface.  
#  
ps=../example_16.ps  
gmtset ANNOT_FONT_SIZE_PRIMARY 9  
#  
pscontour -R0/6.5/-0.2/6.5 -Jx0.45i -P -K -Y5.5i -Ba2f1WSne table_5.11 -Cex16.cpt -I > $ps  
echo ~3.25 7 18 0 4 CB pscontour (triangulate)~ | pstext -R -J -O -K -N >> $ps  
#  
surface table_5.11 -R -I0.2 -Graws0.nc  
grdview raws0.nc -R -J -Ba2f1WSne -Cex16.cpt -Qs -O -K -X3.5i >> $ps  
echo ~3.25 7 18 0 4 CB surface (tension = 0)~ | pstext -R -J -O -K -N >> $ps  
#  
surface table_5.11 -R -I0.2 -Graws5.nc -T0.5  
grdview raws5.nc -R -J -Ba2f1WSne -Cex16.cpt -Qs -O -K -Y-3.75i -X-3.5i >> $ps  
echo ~3.25 7 18 0 4 CB surface (tension = 0.5)~ | pstext -R -J -O -K -N >> $ps  
#  
triangulate table_5.11 -Grawt.nc -R -I0.2 > /dev/null  
grdfilter rawt.nc -Gfiltered.nc -D0 -Fc1  
grdview filtered.nc -R -J -Ba2f1WSne -Cex16.cpt -Qs -O -K -X3.5i >> $ps  
echo ~3.25 7 18 0 4 CB triangulate @~\256@~ grdfilter~ | pstext -R -J -O -K -N >> $ps  
echo ~3.2125 7.5 32 0 4 CB Gridding of Data~ | pstext -R0/10/0/10 -Jx1i -O -K -N -X-3.5i >> $ps  
psscale -D3.25i/0.35i/5i/0.25ih -Cex16.cpt -O -U~Example 16 in Cookbook~ -Y-0.75i >> $ps  
#  
rm -f *.nc .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.16: More ways to grid data.

7.17 Images clipped by coastlines

This example demonstrates how pscoast can be used to set up clip paths based on coastlines. This approach is well suited when different gridded data sets are to be merged on a plot using different color palette files. Merging the files themselves may not be doable since they may represent different data sets, as we show in this example. Here, we lay down a color map of the geoid field near India with grdimage, use pscoast to set up land clip paths, and then overlay topography from the ETOPO5 data set with another call to grdimage. We finally undo the clippath with a second call to pscoast with the option -Q (Figure 7.17):

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 17  
#  
# Purpose:      Illustrates clipping of images using coastlines  
# GMT progs:    grd2cpt, grdgradient, grdimage, pscoast, pstext  
# Unix progs:   rm  
#  
ps=../example_17.ps  
 
# First generate geoid image w/ shading  
 
grd2cpt india_geoid.nc -Crainbow > geoid.cpt  
grdgradient india_geoid.nc -Nt1 -A45 -Gindia_geoid_i.nc  
grdimage india_geoid.nc -Iindia_geoid_i.nc -JM6.5i -Cgeoid.cpt -P -K \  
        -U~Example 17 in Cookbook~ > $ps  
 
# Then use pscoast to initiate clip path for land  
 
pscoast -Rindia_geoid.nc -J -O -K -Dl -Gc >> $ps  
 
# Now generate topography image w/shading  
 
echo ~-10000 150 10000 150~ > gray.cpt  
grdgradient india_topo.nc -Nt1 -A45 -Gindia_topo_i.nc  
grdimage india_topo.nc -Iindia_topo_i.nc -J -Cgray.cpt -O -K >> $ps  
 
# Finally undo clipping and overlay basemap  
 
pscoast -R -J -O -K -Q -B10f5:.~Clipping of Images~: >> $ps  
 
# Put a color legend on top of the land mask  
 
psscale -D4i/7.6i/4i/0.2ih -Cgeoid.cpt -B5f1/:m: -I -O -K >> $ps  
 
# Add a text paragraph  
 
pstext -R -J -O -m -Wwhite,Othinner -D-0.1i/0.1i >> $ps << END  
> 90 -10 12 0 4 RB 12p 3i j  
@_@%5%Example 17.@%%@_  We first plot the color geoid image  
for the entire region, followed by a gray-shaded @#etopo5@#  
image that is clipped so it is only visible inside the coastlines.  
END  
 
# Clean up  
 
rm -f geoid.cpt gray.cpt *_i.nc .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.17: Clipping of images using coastlines.

We also plot a color legend on top of the land. So here we basically have three layers of “paint” stacked on top of each other: the underlaying geoid map, the land mask, and finally the color legend. This legend makes clear how grd2cpt distributed the colors over the range: they are not of equal length put are associated with equal amounts of area in the plot. Since the high amounts (in red) are not very prevalent, that color spans a long range.

For this image it is appropriate to use the -I option in psscale so the legend gets shaded, similar to the geoid grid. See Appendix M to learn more about color palettes and ways to draw color legends.

7.18 Volumes and Spatial Selections

To demonstrate potential usage of the new programs grdvolume and gmtselect we extract a subset of the Sandwell & Smith altimetric gravity field26 for the northern Pacific and decide to isolate all seamounts that (1) exceed 50 mGal in amplitude and (2) are within 200 km of the Pratt seamount. We do this by dumping the 50 mGal contours to disk, then making a simple $AWK script center.awk that returns the mean location of the points making up each closed polygon, and then pass these locations to gmtselect which retains only the points within 200 km of Pratt. We then mask out all the data outside this radius and use grdvolume to determine the combined area and volumes of the chosen seamounts. Our illustration is presented in Figure 7.18.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 18  
#  
# Purpose:      Illustrates volumes of grids inside contours and spatial  
#               selection of data  
# GMT progs:    gmtset, gmtselect, grdclip, grdcontour, grdgradient, grdimage  
# GMT progs:    grdmath, grdvolume, makecpt, pscoast, psscale, pstext, psxy  
# Unix progs:   $AWK, cat, rm  
#  
ps=../example_18.ps  
 
# Use spherical projection since SS data define on sphere  
gmtset ELLIPSOID Sphere D_FORMAT %g  
 
# Define location of Pratt seamount  
echo ~-142.65 56.25~ > pratt.d  
 
# First generate gravity image w/ shading, label Pratt, and draw a circle  
# of radius = 200 km centered on Pratt.  
 
makecpt -Crainbow -T-60/60/10 -Z > grav.cpt  
grdgradient AK_gulf_grav.nc -Nt1 -A45 -GAK_gulf_grav_i.nc  
grdimage AK_gulf_grav.nc -IAK_gulf_grav_i.nc -JM5.5i -Cgrav.cpt -B2f1 -P -K -X1.5i -Y5.85i > $ps  
pscoast -RAK_gulf_grav.nc -J -O -K -Di -Ggray -Wthinnest >> $ps  
psscale -D2.75i/-0.4i/4i/0.15ih -Cgrav.cpt -B20f10/:mGal: -O -K >> $ps  
$AWK ’{print $1, $2, 12, 0, 1, ~LB~, ~Pratt~}’ pratt.d | pstext -R -J -O -K -D0.1i/0.1i >> $ps  
$AWK ’{print $1, $2, 0, 400, 400}’ pratt.d | psxy -R -J -O -K -SE -Wthinnest >> $ps  
 
# Then draw 10 mGal contours and overlay 50 mGal contour in green  
 
grdcontour AK_gulf_grav.nc -J -C20 -B2f1WSEn -O -K -Y-4.85i \  
        -U/-1.25i/-0.75i/~Example 18 in Cookbook~ >> $ps  
grdcontour AK_gulf_grav.nc -J -C10 -L49/51 -O -K -Dsm -Wcthin,green >> $ps  
pscoast -R -J -O -K -Di -Ggray -Wthinnest >> $ps  
$AWK ’{print $1, $2, 0, 400, 400}’ pratt.d | psxy -R -J -O -K -SE -Wthinnest >> $ps  
rm -f sm_*[0-9].xyz     # Only consider closed contours  
 
# Now determine centers of each enclosed seamount > 50 mGal but only plot  
# the ones within 200 km of Pratt seamount.  
 
# First determine mean location of each closed contour and  
# add it to the file centers.d  
 
rm -f centers.d  
for file in sm_*.xyz; do  
        $AWK ’BEGIN{x=0;y=0;n=0};{x+=$1;y+=$2;n++};END{print x/n,y/n}’ $file >> centers.d  
done  
 
# Only plot the ones within 200 km  
 
gmtselect -C200/pratt.d centers.d -fg | psxy -R -J -O -K -SC0.04i -Gred -Wthinnest >> $ps  
psxy -R -J -O -K -ST0.1i -Gyellow -Wthinnest pratt.d >> $ps  
 
# Then report the volume and area of these seamounts only  
# by masking out data outside the 200 km-radius circle  
# and then evaluate area/volume for the 50 mGal contour  
 
grdmath -R ‘$AWK ’{print $1, $2}’ pratt.d‘ SDIST = mask.nc  
grdclip mask.nc -Sa200/NaN -Sb200/1 -Gmask.nc  
grdmath AK_gulf_grav.nc mask.nc MUL = tmp.nc  
area=‘grdvolume tmp.nc -C50 -Sk | cut -f2‘  
volume=‘grdvolume tmp.nc -C50 -Sk | cut -f3‘  
 
psxy -R -J -A -O -K -L -Wthin -Gwhite >> $ps << END  
-148.5  52.75  
-141    52.75  
-141    53.75  
-148.5  53.75  
END  
pstext -R -J -O >> $ps << END  
-148 53.08 14 0 1 LM Areas: $area km@+2@+  
-148 53.42 14 0 1 LM Volumes: $volume mGal\264km@+2@+  
END  
 
# Clean up  
 
rm -f grav.cpt sm_*.xyz *_i.nc tmp.nc mask.nc pratt.d center* .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.18: Volumes and geo-spatial selections.

7.19 Color patterns on maps

GMT 3.1 introduced color patterns and this examples give a few cases of how to use this new feature. We make a phony poster that advertises an international conference on GMT in Honolulu. We use grdmath, makecpt, and grdimage to draw pleasing color backgrounds on maps, and overlay pscoast clip paths to have the patterns change at the coastlines. The middle panel demonstrates a simple pscoast call where the built-in pattern # 86 is drawn at 100 dpi but with the black and white pixels replaced with color combinations. At the same time the ocean is filled with a repeating image of a circuit board (provides in Sun raster format). The text GMT in the center is an off-line PostScript file that was overlaid using psimage. The final panel repeats the top panel except that the land and sea images have changed places (Figure 7.19).

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 19  
#  
# Purpose:      Illustrates various color pattern effects for maps  
# GMT progs:    gmtset, grdimage, grdmath, makecpt, pscoast, pstext, psimage  
# Unix progs:   rm  
#  
ps=../example_19.ps  
 
# First make a worldmap with graded blue oceans and rainbow continents  
 
gmtset COLOR_MODEL rgb  
grdmath -Rd -I1 Y COSD 2 POW = lat.nc  
grdmath -Rd -I1 X Y ABS 90 NEQ MUL = lon.nc  
echo ~0 white 1 blue~ > lat.cpt  
makecpt -Crainbow -T-180/180/60 -Z > lon.cpt  
grdimage lat.nc -Sl -JI0/6.5i -Clat.cpt -P -K -Y7.5i -B0 > $ps  
pscoast -R -J -O -K -Dc -A5000 -Gc >> $ps  
grdimage lon.nc -Sl -J -Clon.cpt -O -K >> $ps  
pscoast -R -J -O -K -Q >> $ps  
pscoast -R -J -O -K -Dc -A5000 -Wthinnest >> $ps  
echo ~0 20 32 0 1 CM 9TH INTERNATIONAL~ | pstext -R -J -O -K -Gred -Sthinner >> $ps  
echo ~0 -10 32 0 1 CM GMT CONFERENCE~ | pstext -R -J -O -K -Gred -Sthinner >> $ps  
echo ~0 -30 18 0 1 CM Honolulu, Hawaii, April 1, 2011~ | pstext -R -J -O -K -Ggreen -Sthinnest >> $ps  
 
# Then show example of color patterns and placing a PostScript image  
 
pscoast -R -J -O -K -Dc -A5000 -Gp100/86:FredByellow -Sp100/circuit.ras -B0 -Y-3.25i >> $ps  
echo ~0 30 32 0 1 CM SILLY USES OF~ | pstext -R -J -O -K -Glightgreen -Sthinner >> $ps  
echo ~0 -30 32 0 1 CM COLOR PATTERNS~ | pstext -R -J -O -K -Gmagenta -Sthinner >> $ps  
psimage -C3.25i/1.625i/CM -W3i GMT_covertext.eps -O -K >> $ps  
 
# Finally repeat 1st plot but exchange the patterns  
 
grdimage lon.nc -Sl -J -Clon.cpt -O -K -Y-3.25i -B0 -U~Example 19 in Cookbook~ >> $ps  
pscoast -R -J -O -K -Dc -A5000 -Gc >> $ps  
grdimage lat.nc -Sl -J -Clat.cpt -O -K >> $ps  
pscoast -R -J -O -K -Q >> $ps  
pscoast -R -J -O -K -Dc -A5000 -Wthinnest >> $ps  
echo ~0 20 32 0 1 CM 9TH INTERNATIONAL~ | pstext -R -J -O -K -Gred -Sthinner >> $ps  
echo ~0 -10 32 0 1 CM GMT CONFERENCE~ | pstext -R -J -O -K -Gred -Sthinner >> $ps  
echo ~0 -30 18 0 1 CM Honolulu, Hawaii, April 1, 2011~ | pstext -R -J -O -Ggreen -Sthinnest >> $ps  
 
rm -f l*.nc l*.cpt .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.19: Using color patterns and additional PostScript material in illustrations.

7.20 Custom plot symbols

One is often required to make special maps that shows the distribution of certain features but one would prefer to use a custom symbol instead of the built-in circles, squares, triangles, etc. in the GMT plotting programs psxy and psxyz. Here we demonstrate one approach that allows for a fair bit of flexibility in designing ones own symbols. The following recipe is used when designing a new symbol.

  1. Use psbasemap (or engineering paper!) to set up an empty grid that goes from -0.5 to +0.5 in both x and y. Use ruler and compass to draw your new symbol using straight lines, arcs of circles, and stand-alone geometrical objects (see psxy man page for a full description of symbol design). In this Section we will create two new symbols: a volcano and a bulls eye.
    PIC

  2. After designing the symbol we will encode it using a simple set of rules. In our case we describe our volcano and bulls eye using these three freeform polygon generators:
    x0 y0 M [ -Gfill ] [ -Wpen ] Start new element at x0, y0

    x1 y1 D Draw straight line from current point to x1, y1 around (x0, y0)

    x0 y0 r α1 α2 ADraw arc segment of radius r from angle α1 to α2

    We also add a few stand-alone circles (for other symbols, see psxy man page):

    x0 y0 r c [ -Gfill ] [ -Wpen ] Draw single circle of radius r around x0, y0

    The optional -G and -W can be used to hardwire the color fill and pen for segments (use to disallow fill or line for any specific feature). By default the segments are painted based on the values of the command line settings.

    Manually applying these rules to our volcano symbol results in a definition file volcano.def:

    ______________________________________________________________________________

    #  
    #       Definition file for a volcano symbol  
    #       To be used with psxy as -Skvolcano/<size>.  
    #       The symbol will be painted and drawn given the  
    #       -G -L -W options on the psxy command line.  
    #  
    -0.5    -0.5    M  
    -0.2    0       D  
    -0.1    0.173205081     0.4     240     300     A  
    0.3     -0.5    D  
    -0.5    -0.5    D  
    -0.05   0.15    0.2     c  
    0.15    0.3     0.15    c  
    0.325   0.4     0.1     c  
    0.45    0.45    0.05    c
    __________________________________________________________________________________________________

    Without much further discussion we also make a definition file bullseye.def for a multi-colored bulls eye symbol. Note that the symbol can be created beyond the -0.5 to +0.5 range, as shown by the red lines. There is no limit in GMT to the size of the symbols. The center, however, will always be at (0,0). This is the point to which the coordinates in psxy refers.

    ______________________________________________________________________________

    #  
    #       Segment info file for bullseye symbol  
    #       These instructions are intended for make_symbol  
    #       which will generate an awk-script that creates  
    #       multiple-segment output describing the desired  
    #       symbol at the chosen size.  The symbol will be  
    #       painted drawn given the -G  -W options for each  
    #       segment.  
    #  
    0       -0.7    M       -W0.5p,red  
    0       0.7     D  
    -0.7    0       M       -W0.5p,red  
    0.7     0       D  
    0       0       0.9     c       -Gp0/12  
    0       0       0.9     c       -W0.25p  
    0       0       0.7     c       -Gyellow -W0.25p  
    0       0       0.5     c       -Gp0/9  
    0       0       0.5     c       -W0.25p  
    0       0       0.3     c       -Gyellow -W0.25p  
    0       0       0.1     c       -Gwhite -W0.25p
    __________________________________________________________________________________________________

    The values refer to positions and dimensions illustrated in the Figure above.

  3. Given proper definition files we may now use them with psxy or psxyz.

We are now ready to give it a try. Based on the hotspot locations in the file hotspots.d (with a 3rd column giving the desired symbol sizes in inches) we lay down a world map and overlay red volcano symbols using our custom-built volcano symbol and psxy. We do something similar with the bulls eye symbols. Without the -G option, however, they get the colors defined in bullseye.def.

Here is our final map script that produces Figure 7.20:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 20  
#  
# Purpose:      Extend GMT to plot custom symbols  
# GMT progs:    pscoast, psxy  
# Unix progs:   rm  
#  
# Plot a world-map with volcano symbols of different sizes  
# on top given locations and sizes in hotspots.d  
ps=../example_20.ps  
 
cat > hotspots.d << END  
55.5    -21.0   0.25  
63.0    -49.0   0.25  
-12.0   -37.0   0.25  
-28.5   29.34   0.25  
48.4    -53.4   0.25  
155.5   -40.4   0.25  
-155.5  19.6    0.5  
-138.1  -50.9   0.25  
-153.5  -21.0   0.25  
-116.7  -26.3   0.25  
-16.5   64.4    0.25  
END  
 
pscoast -Rg -JR9i -B60/30:.~Hotspot Islands and Cities~: -Gdarkgreen -Slightblue -Dc -A5000 -K \  
        -U~Example 20 in Cookbook~ > $ps  
 
psxy -R -J hotspots.d -Skvolcano -O -K -Wthinnest -Gred >> $ps  
 
# Overlay a few bullseyes at NY, Cairo, and Perth  
 
cat > cities.d << END  
286     40.45   0.8  
31.15   30.03   0.8  
115.49  -31.58  0.8  
END  
 
psxy -R -J cities.d -Skbullseye -O >> $ps  
 
rm -f hotspots.d cities.d .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.20: Using custom symbols in GMT.

Given these guidelines you can easily make your own symbols. Symbols with more than one color can be obtained by making several symbol components. E.g., to have yellow smoke coming out of red volcanoes we would make two symbols: one with just the cone and caldera and the other with the bubbles. These would be plotted consecutively using the desired colors. Alternatively, like in bullseye.def, we may specify colors directly for the various segments. Note that the custom symbols (Appendix N), unlike the built-in symbols in GMT, can be used with the built-in patterns (Appendix E). Other approaches are also possible, of course.

7.21 Time-series of RedHat stock price

As discussed in Section 4.4.3, the annotation of time-series is generally more complicated due to the extra degrees of freedom afforded by the dual annotation system. In this example we will display the trend of the stock price of RedHat (RHAT) from their initial public offering until late 2006. The data file is a comma-separated table and the records look like this:

Date,Open,High,Low,Close,Volume,Adj.Close*  
12-Mar-04,17.74,18.49,17.67,18.02,4827500,18.02  
11-Mar-04,17.60,18.90,17.37,18.09,7700400,18.09

Hence, we have a single header record and various prices in USD for each day of business. We will plot the trend of the opening price as a red line superimposed on a yellow envelope representing the low-to-high fluctuation during each day. We also indicate when and at what cost Paul Wessel bought a few shares, and zoom in on the developments since 2004; in the inset we label the time-axis in Finnish in honor of Linus Thorvalds. Because the time coordinates are Y2K-challenged and the order is backwards (big units of years come after smaller units like days) we must change the default input/output formats used by GMT. Finally, we want to prefix prices with the $ symbol to indicate the currency. Here is how it all comes out:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 21  
#  
# Purpose:      Plot a time-series  
# GMT progs:    gmtset, gmtconvert, minmax, psbasemap, psxy  
# Unix progs:   cut, echo  
#  
ps=../example_21.ps  
 
# File has time stored as dd-Mon-yy so set input format to match it  
 
gmtset INPUT_DATE_FORMAT dd-o-yy PLOT_DATE_FORMAT o ANNOT_FONT_SIZE_PRIMARY +10p  
gmtset TIME_FORMAT_PRIMARY abbreviated CHAR_ENCODING ISOLatin1+  
 
# Pull out a suitable region string in yyy-mm-dd format  
 
minmax -fT -I50 -C -H RHAT_price.csv > RHAT.info  
w=‘cut -f1 RHAT.info‘  
e=‘cut -f2 RHAT.info‘  
s=‘cut -f3 RHAT.info‘  
n=‘cut -f4 RHAT.info‘  
R=~-R$w/$e/$s/$n~  
 
# Lay down the basemap:  
 
psbasemap $R -JX9i/6i -Glightgreen -K -U~Example 21 in Cookbook~ -Bs1Y/WSen \  
   -Bpa3Of1o/50WSen:=\$::.~RedHat (RHAT) Stock Price Trend since IPO~: > $ps  
 
# Plot main window with open price as red line over yellow envelope of low/highs  
 
gmtset OUTPUT_DATE_FORMAT dd-o-yy  
gmtconvert -F0,2 -f0T -Hi RHAT_price.csv > RHAT.env  
gmtconvert -F0,3 -f0T -I -Hi RHAT_price.csv >> RHAT.env  
psxy -R -J -Gyellow -O -K RHAT.env >> $ps  
psxy -R -J RHAT_price.csv -H -Wthin,red -O -K >> $ps  
 
# Draw P Wessel’s purchase price as line and label it.  Note we temporary switch  
# back to default yyyy-mm-dd format since that is what minmax gave us.  
 
echo ~05-May-00 0~ > RHAT.pw  
echo ~05-May-00 300~ >> RHAT.pw  
psxy -R -J RHAT.pw -Wthinner,- -O -K >> $ps  
echo ~01-Jan-99 25~ > RHAT.pw  
echo ~01-Jan-07 25~ >> RHAT.pw  
psxy -R -J RHAT.pw -Wthick,- -O -K >> $ps  
gmtset INPUT_DATE_FORMAT yyyy-mm-dd  
echo ~$w 25 12 0 17 LB Wessel purchase price~ | pstext -R -J -O -K -D2i/0.05i -N >> $ps  
gmtset INPUT_DATE_FORMAT dd-o-yy  
 
# Get smaller region for insert for trend since 2004  
 
R=~-R2004T/$e/$s/40~  
 
# Lay down the basemap, using Finnish annotations and place the insert in the upper right:  
 
gmtset TIME_LANGUAGE fi  
psbasemap $R -JX6i/3i -Bpa3Of3o/10:=\$:ESw -Bs1Y/ -Glightblue -O -K -X3i -Y3i >> $ps  
gmtset TIME_LANGUAGE us  
 
# Again, plot close price as red line over yellow envelope of low/highs  
 
psxy -R -J -Gyellow -O -K RHAT.env >> $ps  
psxy -R -J RHAT_price.csv -H -Wthin/red -O -K >> $ps  
 
# Draw P Wessel’s purchase price as dashed line  
 
psxy -R -J RHAT.pw -Wthick,- -O >> $ps  
 
# Clean up after ourselves:  
 
rm -f RHAT.* .gmt*
__________________________________________________________________________________________________

which produces the plot in Figure 7.21, suggesting Wessel has missed a few trains if he had hoped to cash in on the Internet bubble...


PIC

Figure 7.21: Time-series of RedHat stock price since IPO.

7.22 World-wide seismicity the last 7 days

The next example uses the command-line tool wget to obtain a data file from a specified URL27. In the example script this line is commented out so the example will run even if you do not have wget (we use the supplied neic_quakes.d which normally would be created by wget); remove the comment to get the actual current seismicity plot using the live data. The main purpose of this script is not to show how to plot a map background and a few circles, but rather demonstrate how a map legend may be composed using the new tool pslegend. Some scripting is used to pull out information from the data file that is later used in the legend. The legend will normally have the email address of the script owner; here that command is commented out and the user is hardwired to “GMT guru”. The USGS logo, taken from their web page and converted to a Sun raster file, is used to spice up the legend.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 22  
#  
# Purpose:      Automatic map of last 7 days of world-wide seismicity  
# GMT progs:    gmtset, pscoast, psxy, pslegend  
# Unix progs:   cat, sed, awk, wget|curl  
#  
ps=../example_22.ps  
gmtset ANNOT_FONT_SIZE_PRIMARY 10p HEADER_FONT_SIZE 18p PLOT_DEGREE_FORMAT ddd:mm:ssF  
 
# Get the data (-q quietly) from USGS using the wget (comment out in case  
# your system does not have wget or curl)  
 
#wget http://neic.usgs.gov/neis/gis/bulletin.asc -q -O neic_quakes.d  
#curl http://neic.usgs.gov/neis/gis/bulletin.asc -s > neic_quakes.d  
 
# Count the number of events (to be used in title later. one less due to header)  
 
n=‘cat neic_quakes.d | wc -l‘  
n=‘expr $n - 1‘  
 
# Pull out the first and last timestamp to use in legend title  
 
first=‘sed -n 2p neic_quakes.d | $AWK -F, ’{printf ~%s %s\n~, $1, $2}’‘  
last=‘sed -n ’$p’ neic_quakes.d | $AWK -F, ’{printf ~%s %s\n~, $1, $2}’‘  
 
# Assign a string that contains the current user @ the current computer node.  
# Note that two @@ is needed to print a single @ in pstext:  
 
#set me = ~$user@@‘hostname‘~  
me=~GMT guru @@ GMTbox~  
 
# Create standard seismicity color table  
 
cat > neis.cpt << END  
0       red     100     red  
100     green   300     green  
300     blue    10000   blue  
END  
 
# Start plotting. First lay down map, then plot quakes with size = magintude/50~:  
 
pscoast -Rg -JK180/9i -B45g30:.~World-wide earthquake activity~: -Gbrown -Slightblue \  
        -Dc -A1000 -K -U/-0.75i/-2.5i/~Example 22 in Cookbook~ -Y2.75i > $ps  
$AWK -F, ’{ print $4, $3, $6, $5*0.02}’ neic_quakes.d \  
        | psxy -R -JK -O -K -Cneis.cpt -Sci -Wthin -H >> $ps  
# Create legend input file for NEIS quake plot  
 
cat > neis.legend << END  
H 16 1 $n events during $first to $last  
D 0 1p  
N 3  
V 0 1p  
S 0.1i c 0.1i red 0.25p 0.2i Shallow depth (0-100 km)  
S 0.1i c 0.1i green 0.25p 0.2i Intermediate depth (100-300 km)  
S 0.1i c 0.1i blue 0.25p 0.2i Very deep (> 300 km)  
V 0 1p  
D 0 1p  
N 7  
V 0 1p  
S 0.1i c 0.06i - 0.25p 0.3i M 3  
S 0.1i c 0.08i - 0.25p 0.3i M 4  
S 0.1i c 0.10i - 0.25p 0.3i M 5  
S 0.1i c 0.12i - 0.25p 0.3i M 6  
S 0.1i c 0.14i - 0.25p 0.3i M 7  
S 0.1i c 0.16i - 0.25p 0.3i M 8  
S 0.1i c 0.18i - 0.25p 0.3i M 9  
V 0 1p  
D 0 1p  
N 1  
>  
END  
 
# Put together a reasonable legend text, and add logo and user’s name:  
 
cat << END >> neis.legend  
>  
T USGS/NEIS most recent earthquakes for the last seven days.  The data were  
T obtained automatically from the USGS Earthquake Hazards Program page at  
T @_http://neic/usgs.gov @_.  Interested users may also receive email alerts  
T from the USGS.  
T This script can be called daily to update the latest information.  
G 0.4i  
# Add USGS logo  
I USGS.ras 1i RT  
G -0.3i  
L 12 6 LB $me  
END  
 
# OK, now we can actually run pslegend.  We center the legend below the map.  
# Trial and error shows that 1.7i is a good legend height:  
 
pslegend -Dx4.5i/-0.4i/7i/1.7i/TC -J -R -O -F neis.legend -Glightyellow >> $ps  
 
# Clean up after ourselves:  
 
rm -f neis.* .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.22: World-wide seismicity the last 7 days.

The script produces the plot in Figure 7.22, giving the URL where these and similar data can be obtained.

7.23 All great-circle paths lead to Rome

While motorists recently have started to question the old saying “all roads lead to Rome”, aircraft pilots have known from the start that only one great-circle path connects the points of departure and arrival28. This provides the inspiration for our next example which uses grdmath to calculate distances from Rome to anywhere on Earth and grdcontour to contour these distances. We pick five cities that we connect to Rome with great circle arcs, and label these cities with their names and distances (in km) from Rome, all laid down on top of a beautiful world map. Note that we specify that contour labels only be placed along the straight map-line connecting Rome to its antipode, and request curved labels that follows the shape of the contours.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 23  
#  
# Purpose:      Plot distances from Rome and draw shortest paths  
# GMT progs:    grdmath, grdcontour, pscoast, psxy, pstext, grdtrack  
# Unix progs:   echo, cat, awk  
#  
ps=../example_23.ps  
 
# Position and name of central point:  
 
lon=12.50  
lat=41.99  
name=~Rome~  
 
# Calculate distances (km) to all points on a global 1x1 grid  
 
grdmath -Rg -I1 $lon $lat SDIST = dist.nc  
 
# Location info for 5 other cities + label justification  
 
cat << END > cities.d  
105.87  21.02   HANOI           LM  
282.95  -12.1   LIMA            LM  
178.42  -18.13  SUVA            LM  
237.67  47.58   SEATTLE         RM  
28.20   -25.75  PRETORIA        LM  
END  
 
pscoast -Rg -JH90/9i -Glightgreen -Sblue -U~Example 23 in Cookbook~ -A1000 \  
        -B0g30:.~Distances from $name to the World~: -K -Dc -Wthinnest > $ps  
 
grdcontour dist.nc -A1000+v+ukm+kwhite -Glz-/z+ -S8 -C500 -O -K -J \  
        -Wathin,white -Wcthinnest,white,- >> $ps  
 
# For each of the cities, plot great circle arc to Rome with psxy  
 
while read clon clat city; do  
        (echo $lon $lat; echo $clon $clat) | psxy -R -J -O -K -Wthickest/red >> $ps  
done < cities.d  
 
# Plot red squares at cities and plot names:  
psxy -R -J -O -K -Ss0.2 -Gred -Wthinnest cities.d >> $ps  
$AWK ’{print $1, $2, 12, 0, 9, $4, $3}’ cities.d | pstext -R -J -O -K -Dj0.15/0 -Gred -N >> $ps  
# Place a yellow star at Rome  
echo ~$lon $lat~ | psxy -R -J -O -K -Sa0.2i -Gyellow -Wthin >> $ps  
 
# Sample the distance grid at the cities and use the distance in km for labels  
 
grdtrack -Gdist.nc cities.d \  
        | $AWK ’{printf ~%s %s 12 0 1 CT %d\n~, $1, $2, int($NF+0.5)}’ \  
        | pstext -R -J -O -D0/-0.2i -N -Wwhite,o -C0.02i/0.02i >> $ps  
 
# Clean up after ourselves:  
 
rm -f cities.d dist.nc .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.23: All great-circle paths lead to Rome.

The script produces the plot in Figure 7.23; note how interesting the path to Seattle appears in this particular projection (Hammer). We also note that Rome’s antipode lies somewhere near the Chatham plateau (antipodes will be revisited in Section 7.25).

7.24 Data selection based on geospatial criteria

Although we are not seismologists, we have yet another example involving seismicity. We use seismicity data for the Australia/New Zealand region to demonstrate how we can extract subsets of data using geospatial criteria. In particular, we wish to plot the epicenters given in the file oz_quakes.d as red or green circles. Green circles should only be used for epicenters that satisfy the following three criteria:

  1. They are located in the ocean and not on land
  2. They are within 3000 km of Hobart
  3. They are more than 1000 km away from the International Dateline

All remaining earthquakes should be plotted in red. Rather that doing the selection process twice we simply plot all quakes as red circles and then replot those that pass our criteria. Most of the work here is done by gmtselect; the rest is carried out by the usual pscoast and psxy workhorses. Note for our purposes the Dateline is just a line along the 180° meridian.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 24  
#  
# Purpose:      Extract subsets of data based on geospatial criteria  
# GMT progs:    gmtselect, pscoast, psxy, minmax  
# Unix progs:   echo, cat, awk  
#  
# Highlight oceanic earthquakes within 3000 km of Hobart and > 1000 km from dateline  
ps=../example_24.ps  
echo ~147:13 -42:48 3000 Hobart~ > point.d  
cat << END > dateline.d  
> Our proxy for the dateline  
180     0  
180     -90  
END  
R=‘minmax -I10 oz_quakes.d‘  
pscoast $R -JM9i -K -Gtan -Sdarkblue -Wthin,white -Dl -A500 -Ba20f10g10WeSn \  
        -U~Example 24 in Cookbook~ > $ps  
psxy -R -J -O -K oz_quakes.d -Sc0.05i -Gred >> $ps  
gmtselect oz_quakes.d -L1000/dateline.d -Nk/s -C3000/point.d -fg -R -Il \  
        | psxy -R -JM -O -K -Sc0.05i -Ggreen >> $ps  
$AWK ’{print $1, $2, 0, 2*$3, 2*$3}’ point.d | psxy -R -J -O -K -SE -Wfat,white >> $ps  
$AWK ’{print $1, $2, 14, 0, 1, ~LT~, $4}’ point.d | pstext -R -J -O -K -Gwhite -D0.1i/-0.1i >> $ps  
psxy -R -J -O -K point.d -Wfat,white -S+0.2i >> $ps  
psxy -R -J -O -m dateline.d -Wfat,white -A >> $ps  
rm -f point.d dateline.d .gmt*
__________________________________________________________________________________________________

PIC

Figure 7.24: Data selection based on geospatial criteria.

The script produces the plot in Figure 7.24. Note that the horizontal distance from the dateline seems to increase as we go south; however that is just the projected distance (Mercator distortion) and not the actual distance which remains constant at 1000 km.

7.25 Global distribution of antipodes

As promised in Section 7.23, we will study antipodes. The antipode of a point at (ϕ,λ) is the point at (-ϕ,λ +180). We seek an answer to the question that has plagued so many for so long: Given the distribution of land and ocean, how often is the antipode of a point on land also on land? And what about marine antipodes? We use grdlandmask and grdmath to map these distributions and calculate the area of the Earth (in percent) that goes with each of the three possibilities. To make sense of our grdmath equations below, note that we first calculate a grid that is +1 when a point and its antipode is on land, -1 if both are in the ocean, and 0 elsewhere. We then seek to calculate the area distribution of dry antipodes by only pulling out the nodes that equal +1. As each point represent an area approximated by Δϕ ×Δλ where the Δλ term’s actual dimension depends on cos(ϕ), we need to allow for that shrinkage, normalize our sum to that of the whole area of the Earth, and finally convert that ratio to percent. Since the Δλ, Δϕ terms appear twice in these expressions they cancel out, leaving the somewhat intractable expressions below where the sum of cos(ϕ) for all ϕ is known to equal 2Nyπ:

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 25  
#  
# Purpose:      Display distribution of antipode types  
# GMT progs:    gmtset, grdlandmask, grdmath, grd2xyz, gmtmath, grdimage, pscoast, pslegend  
# Unix progs:   cat  
#  
# Create D minutes global grid with -1 over oceans and +1 over land  
ps=../example_25.ps  
D=30  
grdlandmask -Rg -I${D}m -Dc -A500 -N-1/1/1/1/1 -F -Gwetdry.nc  
# Manipulate so -1 means ocean/ocean antipode, +1 = land/land, and 0 elsewhere  
grdmath wetdry.nc DUP 180 ROTX FLIPUD ADD 2 DIV = key.nc  
# Calculate percentage area of each type of antipode match.  
grdmath -Rg -I${D}m -F Y COSD 60 $D DIV 360 MUL DUP MUL PI DIV DIV 100 MUL = scale.nc  
grdmath key.nc -1 EQ 0 NAN scale.nc MUL = tmp.nc  
grd2xyz tmp.nc -S -ZTLf > key.b  
ocean=‘gmtmath -bi1s -Ca -S key.b SUM UPPER RINT =‘  
grdmath key.nc 1 EQ 0 NAN scale.nc MUL = tmp.nc  
grd2xyz tmp.nc -S -ZTLf > key.b  
land=‘gmtmath -bi1s -Ca -S key.b SUM UPPER RINT =‘  
grdmath key.nc 0 EQ 0 NAN scale.nc MUL = tmp.nc  
grd2xyz tmp.nc -S -ZTLf > key.b  
mixed=‘gmtmath -bi1s -Ca -S key.b SUM UPPER RINT =‘  
# Generate corresponding color table  
cat << END > key.cpt  
-1.5    blue    -0.5    blue  
-0.5    gray    0.5     gray  
0.5     red     1.5     red  
END  
# Create the final plot and overlay coastlines  
gmtset ANNOT_FONT_SIZE_PRIMARY +10p PLOT_DEGREE_FORMAT dddF  
grdimage key.nc -Sn -JKs180/9i -B60/30:.~Antipodal comparisons~:WsNE -K -Ckey.cpt -Y1.2i \  
        -U/-0.75i/-0.95i/~Example 25 in Cookbook~ > $ps  
pscoast -R -J -O -K -Wthinnest -Dc -A500 >> $ps  
# Place an explanatory legend below  
pslegend -R0/9/0/0.5 -Jx1i/-1i -O -Dx4.5/0/6i/0.3i/TC -Y-0.2i -Fthick >> $ps << END  
N 3  
S 0.15i s 0.2i red  0.25p 0.3i Terrestrial Antipodes [$land %]  
S 0.15i s 0.2i blue 0.25p 0.3i Oceanic Antipodes [$ocean %]  
S 0.15i s 0.2i gray 0.25p 0.3i Mixed Antipodes [$mixed %]  
END  
rm -f *.nc key.* .gmt*
__________________________________________________________________________________________________

In the end we obtain a funny-looking map depicting the antipodal distribution as well as displaying in legend form the requested percentages (Figure 7.25). Note that the script is set to evaluate a global 30 minute grid for expediency (D = 30), hence several smaller land masses that do have terrestrial antipodes do not show up. If you want a more accurate map you can set the parameter D to a smaller increment (try 5 and wait a few minutes).

The call to grdimage includes the –Sn to suspend interpolation and only return the value of the nearest neighbor. This option is particularly practical for plotting categorical data, like these, that should not be interpolated.


PIC

Figure 7.25: Global distribution of antipodes.

7.26 General vertical perspective projection

Next, we present a recent extension to the -JG projection option which allows the user to specify a particular altitude (this was always at infinity before), as well as several further parameters to limit the view from the chosen vantage point. In this example we show a view of the eastern continental US from a height of 160 km. Below we add a view with a specific tilt of 55° and azimuth 210°; here we have chosen a boresight twist of 45°. We view the land from New York towards Washington, D.C.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 26  
#  
# Purpose:      Demonstrate general vertical perspective projection  
# GMT progs:    pscoast  
# Unix progs:   rm  
#  
ps=../example_26.ps  
 
# first do an overhead of the east coast from 160 km altitude point straight down  
 
latitude=41.5  
longitude=-74.0  
altitude=160.0  
tilt=0  
azimuth=0  
twist=0  
Width=0.0  
Height=0.0  
 
PROJ=-JG${longitude}/${latitude}/${altitude}/${azimuth}/${tilt}/${twist}/${Width}/${Height}/4i  
 
pscoast -Rg $PROJ -X1i -B5g5/5g5 -Glightbrown -Slightblue -W0.25p -Dl -N1/1p,red -N2,0.5p -P -K \  
        -Y5i > $ps  
 
# now point from an altitude of 160 km with a specific tilt and azimuth and with a wider restricted  
# view and a boresight twist of 45 degrees  
 
tilt=55  
azimuth=210  
twist=45  
Width=30.0  
Height=30.0  
 
PROJ=-JG${longitude}/${latitude}/${altitude}/${azimuth}/${tilt}/${twist}/${Width}/${Height}/5i  
 
pscoast -R $PROJ -B5g5/5g5 -Glightbrown -Slightblue -W0.25p -Ia/blue -Di -Na -O -X1i -Y-4i \  
        -U/-1.75i/-0.75i/~Example 26 in Cookbook~ >> $ps  
rm -f .gmt*
__________________________________________________________________________________________________

At this point the full projection has not been properly optimized and the map annotations will need additional work. Also, note that the projection is only implemented in pscoast and grdimage. We hope to refine this further and extend the availability of the full projection to all of the GMT mapping programs.


PIC

Figure 7.26: General vertical perspective projection.

7.27 Plotting Sandwell/Smith Mercator img grids

Next, we show how to plot a data grid that is distributed in projected form. The gravity and predicted bathymetry grids produced by David Sandwell and Walter H. F. Smith are not geographical grids but instead given on a spherical Mercator grid. The GMT supplement imgsrc has tools to extract subsets of these large grids. If you need to make a non-Mercator map then you must extract a geographic grid using img2grd and then plot it using your desired map projection. However, if you want to make a Mercator map then you can save time and preserve data quality by avoiding to re-project the data set twice since it is already in a Mercator projection. This example shows how this is accomplished. We use the -M option in img2grd29 to pull out the grid in Mercator units (i.e., do not invert the Mercator projection) and then simply plot the grid using a linear projection with a suitable scale (here 0.25 inches per degrees of longitude). To overlay basemaps and features that has geographic longitude/latitude coordinates we must remember two key issues:

  1. This is a spherical Mercator grid so we must use –ELLIPSOID=Sphere with all commands that involve projections (or use gmtset to change the setting).
  2. Select Mercator projection and use the same scale that was used with the linear projection.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 27  
#  
# Purpose:      Illustrates how to plot Mercator img grids  
# GMT progs:    makecpt, grdgradient, grdimage, grdinfo, pscoast  
# GMT supplement: img2grd (to read Sandwell/Smith img files)  
# Unix progs:   rm, grep, $AWK  
#  
ps=../example_27.ps  
 
# First extract a chunk of faa and retain short int precision to  
# save disk space.  Gravity is thus in 0.1 mGal increments.  
# Next get gradients.  The grid’s region is in Mercator x/y units  
 
#img2grd grav.15.2.img -R145/170/-50/-25 -M -C -T1 -Gtasman_grav.nc=ns  
grdgradient tasman_grav.nc -Nt1 -A45 -Gtasman_grav_i.nc  
 
# Make a suitable cpt file for mGal  
 
makecpt -T-120/120/10 -Z -Crainbow > grav.cpt  
 
# Since this is a Mercator grid we use a linear projection  
 
grdimage tasman_grav.nc=ns/0.1 -Itasman_grav_i.nc -Jx0.25i -Cgrav.cpt -P -K \  
        -U~Example 27 in Cookbook~ > $ps  
 
# Then use pscoast to plot land; get original -R from grid remark  
# and use Mercator projection with same scale as above on a spherical Earth  
 
R=‘grdinfo tasman_grav.nc | grep Remark | $AWK ’{print $NF}’‘  
 
pscoast $R -Jm0.25i -Ba10f5WSne -O -K -Gblack --ELLIPSOID=Sphere \  
        -Cwhite -Dh+ --PLOT_DEGREE_FORMAT=dddF >> $ps  
 
# Put a color legend on top of the land mask justified with 147E,31S  
 
echo 147E 31S | mapproject -R -J -Di --ELLIPSOID=Sphere > tmp  
echo 147E 31S 1 2.5 | psxy -R -J -O -K -Sr -D0.25i/0.05i -Gwhite -W1p --ELLIPSOID=Sphere --MEASURE_UNIT=inch >> $ps  
pos=‘$AWK ’{printf ~%si/%si\n~, $1, $2}’ tmp‘  
psscale -D$pos/2i/0.15i -Cgrav.cpt -B50f10/:mGal: -I -O >> $ps  
 
# Clean up  
 
rm -f grav.cpt *_i.nc .gmt* tmp
__________________________________________________________________________________________________

This map of the Tasman Sea shows the marine gravity anomalies with land painted black. A color scale bar was then added to complete the illustration.


PIC

Figure 7.27: Plotting Sandwell/Smith Mercator img grids.

7.28 Mixing UTM and geographic data sets

Next, we present a similar case: We wish to plot a data set given in UTM coordinates and want it to be properly registered with overlying geographic data, such as coastlines or data points. The mistake many GMT rookies make is to specify the UTM projection with their UTM data. However, that data have already been projected and is now in linear meters. The only sensible way to plot such data is with a linear projection, yielding a UTM map. In this step one can choose to annotate or tick the map in UTM meters as well. To plot geographic (lon/lat) data on the same map there are a few things you must consider:

  1. You need to know the lower left and upper right UTM coordinates of your map. Given the UTM zone you can use mapproject to recover the lon/lat of those two points. Conversely, if you instead know the lon/lat corners then you need to convert those to UTM coordinates. You now have the ability to specify two domains with the -R setting: The linear UTM meter domain when plotting UTM data and the geographic domain (remember to use the rectangular variant of -R that ends with the modifier r) when plotting lon/lat data.
  2. Make sure you use the same scale (and not width) with both the linear and UTM projection.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 28  
#  
# Purpose:      Illustrates how to mix UTM data and UTM projection  
# GMT progs:    makecpt, grdgradient, grdimage, grdinfo, pscoast, pstext, mapproject  
# Unix progs:   rm, cut, grep, $AWK  
#  
ps=../example_28.ps  
 
# Get intensity grid and set up a color table  
grdgradient Kilauea.utm.nc -Nt1 -A45 -GKilauea.utm_i.nc  
makecpt -Ccopper -T0/1500/100 -Z > Kilauea.cpt  
# Save min/max UTM coordinates with enough precision  
grdinfo Kilauea.utm.nc --D_FORMAT=%.10g -C > tmp.txt  
# Use inverse UTM projection to determine the lon/lat of the lower left and upper right corners  
LL=‘cut -f2,4 tmp.txt | mapproject -Ju5Q/1:1 -F -C -I --OUTPUT_DEGREE_FORMAT=ddd:mm:ss.x | \  
        $AWK ’{printf ~%s/%s\n~, $1, $2}’‘  
UR=‘cut -f3,5 tmp.txt | mapproject -Ju5Q/1:1 -F -C -I --OUTPUT_DEGREE_FORMAT=ddd:mm:ss.x | \  
        $AWK ’{printf ~%s/%s\n~, $1, $2}’‘  
# Lay down the UTM topo grid using a 1:17,000 scale  
grdimage Kilauea.utm.nc -IKilauea.utm_i.nc -CKilauea.cpt -Jx1:170000 -P -K -B5000g5000WSne \  
        -U~Example 28 in Cookbook~ --D_FORMAT=%.10g --ANNOT_FONT_SIZE_PRIMARY=9 \  
        --GRID_CROSS_SIZE_PRIMARY=0.1i > $ps  
# Overlay geographic data and coregister by using correct region and projection with the same scale  
pscoast -R$LL/${UR}r -Ju5Q/1:170000 -O -K -Df+ -Slightblue -W0.5p -B5mg5mNE \  
        --ANNOT_FONT_SIZE_PRIMARY=12 --PLOT_DEGREE_FORMAT=ddd:mmF >> $ps  
psbasemap -R -J -O -K --ANNOT_FONT_SIZE_PRIMARY=9 -Lf155:07:30W/19:15:40N/19:23N/5k+l1:17,000+u \  
        --LABEL_FONT_SIZE=10 >> $ps  
echo 155:16:20W 19:26:20N 12 0 1 CB KILAUEA | pstext -R -J -O >> $ps  
# Clean up  
 
rm -f Kilauea.utm_i.nc Kilauea.cpt tmp.txt .gmt*
__________________________________________________________________________________________________

Our script illustrates how we would plot a UTM grid of elevations near Kilauea volcano on the Big Island of Hawaii. Given we are in UTM zone 5Q, the script determines the geographic coordinates of the lower left and upper right corner of the UTM grid, then uses that region when overlaying the coastline and light blue ocean. We place a scale bar and label Kilauea crater to complete the figure.


PIC

Figure 7.28: Mixing UTM and geographic data sets requires knowledge of the map region domain in both UTM and lon/lat coordinates and consistent use of the same map scale.

7.29 Gridding spherical surface data using splines

Next, we demonstrate how gridding on a spherical surface can be accomplished using Green’s functions of surface splines, with or without tension. Global gridding does not work particularly well in Cartesian coordinates hence the chosen approach. We use greenspline to produce a crude topography grid for Mars based on radii estimates from the Mariner 9 and Viking Orbiter spacecrafts. This data comes from Smith and Zuber [Science, 1996] and is used here as a small (N = 370) data set we can use to demonstrate spherical surface gridding. Since greenspline must solve a N by N matrix system your system memory may impose limits on how large data sets you can handle; also note that the spherical surface spline in tension is particularly slow to compute.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 29  
#  
# Purpose:      Illustrates spherical surface gridding with Green’s function of splines  
# GMT progs:    makecpt, grdcontour, grdgradient, grdimage, grdmath greenspline, psscale, pstext  
# Unix progs:   rm, echo  
#  
ps=../example_29.ps  
 
# This example uses 370 radio occultation data for Mars to grid the topography.  
# Data and information from Smith, D. E., and M. T. Zuber (1996), The shape of  
# Mars and the topographic signature of the hemispheric dichotomy, Science, 271, 184–187.  
 
# Make Mars ellipsoid given their three best-fitting axes:  
a=3399.472  
b=3394.329  
c=3376.502  
grdmath -Rg -I4 -F X COSD $a DIV DUP MUL X SIND $b DIV DUP MUL ADD Y COSD DUP MUL MUL Y SIND $c DIV \  
        DUP MUL ADD SQRT INV = ellipsoid.nc  
 
#  Do both Parker and Wessel/Becker solutions (tension = 0.9975)  
greenspline -Rellipsoid.nc mars370.in -D4 -Sp -Gmars.nc  
greenspline -Rellipsoid.nc mars370.in -D4 -SQ0.9975/5001 -Gmars2.nc  
# Scale to km and remove ellipsoid  
grdmath mars.nc 1000 DIV ellipsoid.nc SUB = mars.nc  
grdmath mars2.nc 1000 DIV ellipsoid.nc SUB = mars2.nc  
makecpt -Crainbow -T-7/15/1 -Z > mars.cpt  
grdgradient mars2.nc -M -Ne0.75 -A45 -Gmars2_i.nc  
grdimage mars2.nc -Imars2_i.nc -Cmars.cpt -B30g30Wsne -JH0/6i -P -K -Ei \  
        -U~Example 29 in Cookbook~ --ANNOT_FONT_SIZE_PRIMARY=12 > $ps  
grdcontour mars2.nc -J -O -K -C1 -A5 -Glz+/z- >> $ps  
psxy -Rg -J -O -K -Sc0.045i -Gblack mars370.in  >> $ps  
echo ~0 90 14 0 1 LB b)~ | pstext -R -J -O -K -N -D-3i/-0.2i >> $ps  
grdgradient mars.nc -M -Ne0.75 -A45 -Gmars_i.nc  
grdimage mars.nc -Imars_i.nc -Cmars.cpt -B30g30Wsne -J -O -K -Y3.6i -Ei --ANNOT_FONT_SIZE_PRIMARY=12 >> $ps  
grdcontour mars.nc -J -O -K -C1 -A5 -Glz+/z- >> $ps  
psxy -Rg -J -O -K -Sc0.045i -Gblack mars370.in  >> $ps  
psscale -Cmars.cpt -O -K -D3i/-0.1i/5i/0.1ih -I --ANNOT_FONT_SIZE_PRIMARY=12 -B2f1/:km: >> $ps  
echo ~0 90 14 0 1 LB a)~ | pstext -R -J -O -N -D-3i/-0.2i >> $ps  
# Clean up  
rm -f *.nc mars.cpt .gmt*
__________________________________________________________________________________________________

Our script must first estimate the ellipsoidal shape of Mars from the parameters given by Smith and Zuber so that we can remove this reference surface from the gridded radii. We run the gridding twice: First with no tension using Parker’s [1990] method and then with tension using the Wessel and Becker [2008] method. The grids are then imaged with grdimage and grdcontour and a color scale is placed between them.


PIC

Figure 7.29: Gridding of spherical surface data using Green’s function splines.

7.30 Trigonometric functions plotted in graph mode

Finally, we end with a simple mathematical illustration of sine and cosine, highlighting the it graph mode for linear projections and the new curved vectors for angles.

_________________________________________________________________________________

#!/bin/bash  
#               GMT EXAMPLE 30  
#  
# Purpose:      Show graph mode and math angles  
# GMT progs:    gmtmath, psbasemap, pstext and psxy  
# Unix progs:   echo, rm  
#  
# Draw generic x-y axes with arrows  
ps=../example_30.ps  
 
psbasemap -R0/360/-1.25/1.75 -JX8i/6i -B90f30:,-\\312:/1g10:.~Two Trigonometric Functions~:WS -K \  
        -U~Example 30 in Cookbook~ --BASEMAP_TYPE=graph --VECTOR_SHAPE=0.5 > $ps  
 
# Draw sine an cosine curves  
 
gmtmath -T0/360/0.1 T COSD = | psxy -R -J -O -K -W2p >> $ps  
gmtmath -T0/360/0.1 T SIND = | psxy -R -J -O -K -W2p,. --PS_LINE_CAP=round >> $ps  
 
# Indicate the x-angle = 120 degrees  
psxy -R -J -O -K -W0.5p,- << EOF >> $ps  
120     -1.25  
120     1.25  
EOF  
 
pstext -R -J -O -K -Dj0.05i -N << EOF >> $ps  
360 1 18 0 4 RB x = cos(@%12%a@%%)  
360 0 18 0 4 RB y = sin(@%12%a@%%)  
120 -1.25 14 0 4 LB 120\\312  
370 -1.35 24 0 12 LT a  
-5 1.85 24 0 4 RT x,y  
EOF  
 
# Draw a circle and indicate the 0-70 degree angle  
 
echo 0 0 | psxy -R-1/1/-1/1 -Jx1.5i -O -K -X3.625i -Y2.75i -Sc2i -W1p -N >> $ps  
psxy -R -J -O -K -m -W1p << EOF >> $ps  
> x-gridline  -W0.25p  
-1      0  
1       0  
> y-gridline  -W0.25p  
0       -1  
0       1  
> angle = 0  
0       0  
1       0  
> angle = 120  
0       0  
-0.5    0.866025  
> x-projection -W2p  
-0.3333 0  
0       0  
> y-projection -W2p  
-0.3333 0.57735  
-0.3333 0  
EOF  
 
pstext -R -J -O -K -Dj0.05i << EOF >> $ps  
-0.16666 0 12 0 4 CT x  
-0.3333 0.2888675 12 0 4 RM y  
0.22 0.27 12 -30 12 CB a  
-0.33333 0.6 12 30 4 LB 120\\312  
EOF  
 
echo 0 0 0 120 | psxy -R -J -O -Sml1i -W1p >> $ps  
 
rm -f .gmt*
__________________________________________________________________________________________________

The script simply draws a graph basemap, computes sine and cosine and plots them as lines, then indicates on a circle that these quantities are simply the projections of an unit vector on the x- and y-axis, at the given angle.


PIC

Figure 7.30: Trigonometric functions plotted in graph mode.

Chapter 8
Creating GMT Animations

Unlike the previous chapter, in this chapter we will explore what is involved in creating animations (i.e., movies). Of course, an animation is nothing more than a series of individual images played back in an orderly fashion. Here, these images will have been created with GMT. To ensure a smooth transition from frame to frame we will be following some general guidelines when writing our scripts. Since there is no “movie” mode in GMT we must take care of all the book-keeping in our script. Thus, animations may require a bit of planning and may use more advanced scripting than the previous static examples. Note: This is a new chapter introduced with the 4.4.0 version and should be considered work in progress.

Most, if not all, animation scripts must deal with several specific phases of movie making:

  1. Define parameters that determine the dimension of the final movie.
  2. Pre-calculate all variables, data tables, grids, or background map layers that are independent of your time variable.
  3. Have a frame-number loop where each frame is created as a PostScript plot, then rasterized to a TIFF file of chosen dimension.
  4. Convert the individual frames to a single movie of suitable format.
  5. Clean up temporary files and eventually the individual frames.

We will discuss these phases in more detail before showing our first example.

1.
There are several coordinates that you need to consider when planning your movie. The first is the coordinates of your data, i.e., the user coordinates. As with all GMT plots you will transform those to the second set of plot coordinates in inches (or cm) by applying a suitable region and map projection. As before, you normally do this with a particular paper size in mind. When printed you get a high-resolution plot in monochrome or color. However, movies are not device-independent and you must finally consider the third set of pixel coordinates which specifies the resolution of the final movie. We control the frame size by selecting a suitable dpi setting that will scale your physical dimensions to the desired frame size in pixels. If you decide up front on a particular resolution (e.g., 480 by 320 pixels) then you should specify a paper size and dpi so that their product yields the desired pixel dimensions. For instance, here it might make sense to plan your plotting on a 4.8 by 3.2 inch “paper” and use 100 dpi to convert it to pixels, but you are free to use any combination that multiplies to the desired dimensions. After deciding on frame size you need to consider how many frames your movie should have. This depends on lots of things such as how patient you are, how many frames per second you need and the time range of your animation. We recommend you use variables to specify the items that go into computing the number of frames so that you can easily test your script with a few frames before changing settings and running the full Hollywood production overnight.
2.
Depending on what you want to display, there are usually many elements that do not change between frames. Examples include a coastline base map for background, an overlay of text legends, perhaps some variables that hold information that will be used during the movie, and possibly subsets of larger data sets. Since movie-making can take a long time if you are ambitious, it is best to compute or plot all the elements that can be done outside your main frame-loop rather than waste time doing the same thing over and over again. You are then ready for the main loop.
3.
Initialize a frame counter to 0 and have a loop that continues until your frame counter equals the desired number of frames. You must use your frame counter to create a unique file name for each frame image so that the series of images can be lexically arranged. We recommend using the GMT shell function gmt_set_framename to format the frame counter with an adequate number of leading zeros; see our examples for details. The bulk of your main loop involves create the single PostScript plot for this particular frame (time). This can be trivial or a serious scripting exercise depending on what you want to show. We will give a few examples with increasing complexity. Once the PostScript plot is created you need to rasterize it; we recommend you use ps2raster to generate a TIFF image at the agreed-upon resolution. We also recommend that you place all frame images in a sub-directory. You may increment your frame counter using gmt_set_framenext.
4.
Once you have all your frames you are ready to combine them into an animation. There are two general approaches. (a) If your image sequence is not too long then you can convert the images into a single animated GIF file. This file can be included in PowerPoint presentations or placed on a web page and will play back as a movie by pausing the specified amount between frames, optionally repeating the entire sequence one or more times. (b) For more elaborate projects you will need to convert the frames into a proper movie format such as MPEG-4. There are both free and commercial tools that can help with this conversion and they tend to be platform-specific. For moderate projects you can use convert to build a MP4 file, but longer movies will require a more flexible tools, such as ffmpeg. These solutions are preferred as they can be scripted and be included in your animation script. Most commercial movie tools such as iMovie or MovieMaker can ingest still images and let you specify the frame duration. Under OS X we prefer to use QuickTime 7.30 Free tools exist to call the Quicktime library functions from the command line as we prefer to do in our scripts. You will find yourself experimenting with compression settings and movie formats so that the final movie has the resolution and portability you require.
5.
Finally, when all is done you should delete any temporary files created. However, since creating the frames may take a lot of time it is best to not automatically delete the frame sub directory. That way you can redo the frames-to-movie conversion with different settings until you are satisfied.

Note that in our examples below the animation scripts only produce a PostScript plot of the first frame, then exits. This is to allow our install script to just produce the frame for inclusion in our documentation, and to avoid a lengthy movie production when installing GMTṪo actually build these animation examples you must run the scripts with an argument (any argument, it does not matter).

8.1 Animation of the sine function

Our first animation is not very ambitious: We wish to plot the sine function from 0–360° and take snap shots every 20°. To get a smooth curve we must sample the function much more frequently; we settle on 10 times more frequently than the frame spacing. We place a bright red circle at the leading edge of the curve, and as we move forward in time (here, angles) we dim the older circles to a dark red color. We add a label that indicates the current angle value. Once the 18 frames are completed we convert them to a single animated GIF file and write a plain HTML wrapper with a simple legend. Opening the HTML page anim01.html in the browser will display the animation.

_________________________________________________________________________________

#!/bin/bash  
#               GMT ANIMATION 01  
#  
# Purpose:      Make web page with simple animated GIF of sine function  
# GMT progs:    gmtset, gmtmath, psbasemap, pstext, psxy, ps2raster  
# Unix progs:   awk, mkdir, rm, mv, echo, convert, cat  
# Note:         Run with any argument to build movie; otherwise 1st frame is plotted only.  
#  
# 1. Initialization  
# 1a) Assign movie parameters  
width=4i  
height=2i  
dpi=125  
n_frames=18  
name=‘basename $0 ’.sh’‘  
# 1b) Do frame-independent calculations and setup  
angle_step=‘gmtmath -Q 360 $n_frames DIV =‘  
angle_inc=‘gmtmath -Q $angle_step 10 DIV =‘  
gmtset DOTS_PR_INCH $dpi  
psbasemap -R0/360/-1.2/1.6 -JX3.5i/1.65i -P -K -X0.35i -Y0.25i \  
        -Ba90g90f30:,-\\312:/a0.5f0.1g1WSne -Glightgreen \  
        --PAPER_MEDIA=Custom_${width}x${height} --ANNOT_FONT_SIZE=+9p > $$.map.ps  
# 2. Main frame loop  
mkdir -p $$  
frame=0  
while [ $frame -le $n_frames ]; do  
        # Create file name using a name_##.tif format  
        file=‘gmt_set_framename $name $frame‘  
        cp -f $$.map.ps $$.ps  
        angle=‘gmtmath -Q $frame $angle_step MUL =‘  
        if [ $frame -gt 0 ]; then       # First plot has no curves  
#               Plot smooth blue curve and dark red dots at all angle steps so far  
                gmtmath -T0/$angle/$angle_inc T SIND = $$.sin.d  
                psxy -R -J -O -K -W1p,blue $$.sin.d >> $$.ps  
                gmtmath -T0/$angle/$angle_step T SIND = $$.sin.d  
                psxy -R -J -O -K -Sc0.1i -Gdarkred $$.sin.d >> $$.ps  
        fi  
        #       Plot red dot at current angle and annotate  
        sin=‘gmtmath -Q $angle SIND =‘  
        echo $angle $sin | psxy -R -J -O -K -Sc0.1i -Gred >> $$.ps  
        echo $angle | awk ’{printf ~0 1.6 14 0 1 LT a = %3.3d\n~, $1}’ \  
                | pstext -R -J -O -K -N -Dj0.1i/0.05i >> $$.ps  
        psxy -R -J -O -T >> $$.ps  
        if [ $# -eq 0 ]; then  
                mv $$.ps ../$name.ps  
                gmt_cleanup .gmt  
                gmt_abort ~$0: First frame plotted to $name.ps~  
        fi  
#       RIP to TIFF at specified dpi  
        ps2raster -E$dpi -Tt $$.ps  
        mv -f $$.tif $$/$file.tif  
        echo ~Frame $file completed~  
        frame=‘gmt_set_framenext $frame‘  
done  
# 3. Create animated GIF file and HTML for web page  
convert -delay 20 -loop 0 $$/*.tif $name.gif  
cat << END > $name.html  
<HTML>  
<TITLE>GMT Trigonometry: The sine movie</TITLE>  
<BODY bgcolor=~#ffffff~>  
<CENTER>  
<H1>GMT Trigonometry: The sine movie</H1>  
<IMG src=~$name.gif~>  
</CENTER>  
<HR>  
We demonstrate how the sine function <I>y = sin(a)</I> varies with <I>a</I> over  
the full 360-degree interval.  We plot a bright red circle at each  
new angle, letting previous circles turn dark red.  The underlying  
sine curve is sampled at 10 times the frame sampling rate in order to reproduce  
a smooth curve.  Our animation uses Imagemagick’s convert tool to make an animated GIF file  
with a 0.2 second pause between frames, set to repeat forever.  
<HR>  
<I>$name.sh: Created by $USER on ‘date‘</I>  
</BODY>  
</HTML>  
END  
# 4. Clean up temporary files  
gmtset DOTS_PR_INCH 300  
gmt_cleanup .gmt
__________________________________________________________________________________________________

Make sure you understand the purpose of all the steps in our script. In this case we did some trial-and-error to determine the exact values to use for the map projection, the region, the spacing around the frame, etc. so that the final result gave a reasonable layout. Do this planning on a single PostScript plot before running a lengthy animation script.


PIC

Figure 8.1: Animation of a simple sine function.

8.2 Examining DEMs using variable illumination

Our next animation uses a gridded topography for parts of Colorado (US); the file is distributed with the tutorial examples. Here, we want to use grdimage to generate a shaded-relief image sequence in which we sweep the illumination azimuth around the entire horizon. The resulting animation illustrates how changing the illumination azimuth can bring out subtle features (or artifacts) in the gridded data. The red arrow points in the direction of the illumination.

_________________________________________________________________________________

#!/bin/bash  
#               GMT ANIMATION 02  
#  
# Purpose:      Make web page with simple animated GIF of a DEM grid  
# GMT progs:    gmtset, gmtmath, grdgradient, makecpt, grdimage psxy, ps2raster  
# Unix progs:   awk, mkdir, rm, mv, echo, convert, cat  
# Note:         Run with any argument to build movie; otherwise 1st frame is plotted only.  
#  
# 1. Initialization  
# 1a) Assign movie parameters  
width=3.5i  
height=4.15i  
dpi=72  
n_frames=36  
TDIR=../../tutorial  
name=‘basename $0 ’.sh’‘  
# 1b) setup  
del_angle=‘gmtmath -Q 360 $n_frames DIV =‘  
makecpt -Crainbow -T500/4500/500 -Z > $$.cpt  
gmtset DOTS_PR_INCH $dpi  
R=‘gmt_get_gridregion $TDIR/us.nc‘  
# 2. Main loop  
mkdir -p $$  
frame=0  
while [ $frame -lt $n_frames ]; do  
        # Create file name using a name_##.tif format  
        file=‘gmt_set_framename $name $frame‘  
        angle=‘gmtmath -Q $frame $del_angle MUL =‘  
        dir=‘gmtmath -Q $angle 180 ADD =‘  
        grdgradient $TDIR/us.nc -A$angle -Nt2 -M -G$$.us_int.nc  
        grdimage $TDIR/us.nc -I$$.us_int.nc -JM3i -P -K -C$$.cpt -B1WSne -X0.35i -Y0.3i \  
        --PAPER_MEDIA=Custom_${width}x${height} --ANNOT_FONT_SIZE=+9p > $$.ps  
        echo 256.25 35.6 | psxy -R$R -J -O -K -Sc0.8i -Gwhite -Wthin >> $$.ps  
        echo 256.25 35.6 $dir 0.37 | psxy -R$R -J -O -Sv0.02i/0.05i/0.05i -Gred -Wthin >> $$.ps  
        if [ $# -eq 0 ]; then  
                mv $$.ps ../$name.ps  
                gmt_cleanup .gmt  
                gmt_abort ~$0: First frame plotted to $name.ps~  
        fi  
#       RIP to TIFF at specified dpi  
        ps2raster -E$dpi -Tt $$.ps  
        mv -f $$.tif $$/$file.tif  
        echo ~Frame $file completed~  
        frame=‘gmt_set_framenext $frame‘  
done  
# 3. Create animated GIF file and HTML for web page  
convert -delay 10 -loop 0 $$/*.tif $name.gif  
cat << END > $name.html  
<HTML>  
<TITLE>GMT shading: A tool for feature detection</TITLE>  
<BODY bgcolor=~#ffffff~>  
<CENTER>  
<H1>GMT shading: A tool for feature detection</H1>  
<IMG src=~$name.gif~>  
</CENTER>  
<HR>  
We make illuminated images of topography from a section of Colorado and  
vary the azimuth of the illumination (see arrow).  As the light-source sweeps around  
the area over 360 degrees we notice that different features in the data  
become hightlighted.  This is because the illumination is based on data  
gradients and such derivatives will high-light short-wavelength signal.  
Again, our animation uses Imagemagick’s convert tool to make an animated GIF file  
with a 0.1 second pause between the 36 frames.  
<HR>  
<I>$name.sh: Created by $USER on ‘date‘</I>  
</BODY>  
</HTML>  
END  
# 4. Clean up temporary files  
gmtset DOTS_PR_INCH 300  
gmt_cleanup .gmt
__________________________________________________________________________________________________

As you can see, these sorts of animations are not terribly difficult to put together, especially since our vantage point is fixed. In the next example we will move the “camera” around and must therefore deal with how to frame perspective views.


PIC

Figure 8.2: Animation of a DEM using variable illumination.

8.3 Orbiting a static map

Our third animation keeps a fixed gridded data set but moves the camera angle around the full 360°. We use grdview to generate a shaded-relief image sequence using the new enhanced -E option. No additional information is plotted on the image. As before we produce an animated GIF image and a simple HTML wrapper for it.

_________________________________________________________________________________

#!/bin/bash  
#               GMT ANIMATION 03  
#  
# Purpose:      Make web page with simple animated GIF of Iceland topo  
# GMT progs:    gmtset, gmtmath, psbasemap, pstext, psxy, ps2raster  
# Unix progs:   awk, mkdir, rm, mv, echo, convert, cat  
# Note:         Run with any argument to build movie; otherwise 1st frame is plotted only.  
#  
# 1. Initialization  
# 1a) Assign movie parameters  
lon=-20  
lat=65  
dpi=100  
x0=1.5  
y0=0.75  
px=4  
py=2.5  
el=35  
az=0  
name=‘basename $0 ’.sh’‘  
mkdir -p $$  
gmtset DOTS_PR_INCH $dpi  
frame=0  
grdclip -Sb0/-1 -G$$_above.nc Iceland.nc  
grdgradient -M -A45 -Nt1 $$_above.nc -G$$.nc  
makecpt -Crelief -Z > $$.cpt  
while [ $az -lt 360 ]; do  
        file=‘gmt_set_framename $name $frame‘  
        if [ $# -eq 0 ]; then   # If a single frame is requested we pick this view  
                az=135  
        fi  
        grdview $$_above.nc -JM2.5i -C$$.cpt -Qi$dpi -B5g10/5g5 -E$az/${el}+w$lon/${lat}+v$x0/$y0 -P -X0.5i -Y0.5i --PAPER_MEDIA=Custom_${px}ix${py}i > $$.ps  
        if [ $# -eq 0 ]; then  
                mv $$.ps ../$name.ps  
                gmt_cleanup .gmt  
                gmt_abort ~$0: First frame plotted to $name.ps~  
        fi  
        ps2raster $$.ps -Tt -E$dpi  
        mv $$.tif $$/$file.tif  
        az=‘expr $az + 5‘  
        echo ~Frame $file completed~  
        frame=‘gmt_set_framenext $frame‘  
done  
convert -delay 10 -loop 0 $$/*.tif $name.gif  
cat << END > $name.html  
<HTML>  
<TITLE>GMT 3-D perspective of Iceland</TITLE>  
<BODY bgcolor=~#ffffff~>  
<CENTER>  
<H1>GMT 3-D perspective of Iceland</H1>  
<IMG src=~$name.gif~ border=1>  
</CENTER>  
<HR>  
Here we show ETOPO2 topography of Iceland as we move the view  
point around the island.  
<I>$name.sh: Created by $USER on ‘date‘</I>  
</BODY>  
</HTML>  
END  
# 4. Clean up temporary files  
gmtset DOTS_PR_INCH 300  
gmt_cleanup .gmt
__________________________________________________________________________________________________

PIC

Figure 8.3: Orbiting a static map.

8.4 Flying over topography

Our next animation simulates what an imaginary satellite might see as it passes in a great circle from New York to Miami at an altitude of 160 km. We use the general perspective view projection with grdimage and use project to create a great circle path between the two cities, sampled every 5 km. The main part of the script will make the DVD-quality frames from different view points, draw the path on the ground, and add frame numbers to each frame. As this animation generates 355 frames we can use 3rd party tools to turn the image sequence into a MPEG-4 movie31. Note: At the moment, grdview cannot use general perspective view projection to allow “fly-through” animations like Fledermaus; we expect to add this functionality in a future version.

_________________________________________________________________________________

#!/bin/bash  
#               GMT ANIMATION 04  
#  
# Purpose:      Make DVD-res MP4 movie of NY to Miami flight  
# GMT progs:    gmtset, gmtmath, psbasemap, pstext, psxy, ps2raster  
# Unix progs:   awk, mkdir, rm, mv, echo, qt_export, cat  
# Note:         Run with any argument to build movie; otherwise 1st frame is plotted only.  
#  
# 1. Initialization  
# 1a) Assign movie parameters  
REGION=-Rg  
altitude=160.0  
tilt=55  
azimuth=210  
twist=0  
Width=36.0  
Height=34.0  
px=7.2  
py=4.8  
dpi=100  
name=‘basename $0 ’.sh’‘  
 
# Set up flight path  
project -C-73.8333/40.75 -E-80.133/25.75 -G5 -Q > $$.path.d  
frame=0  
mkdir -p frames  
grdgradient USEast_Coast.nc -A90 -Nt1 -G$$_int.nc  
makecpt -Cglobe -Z > $$.cpt  
while read lon lat dist; do  
        file=‘gmt_set_framename $name $frame‘  
        ID=‘echo $frame | awk ’{printf ~%4.4d\n~, $1}’‘  
        grdimage -JG${lon}/${lat}/${altitude}/${azimuth}/${tilt}/${twist}/${Width}/${Height}/7i+ \  
                $REGION -P -Y0.1i -X0.1i USEast_Coast.nc -I$$_int.nc -C$$.cpt \  
                --PAPER_MEDIA=Custom_${px}ix${py}i -K > $$.ps  
        psxy -R -J -O -K -W1p $$.path.d >> $$.ps  
        echo 0 4.6 14 0 1 TL $ID | pstext -R0/$px/0/$py -Jx1i -O >> $$.ps  
        if [ $# -eq 0 ]; then  
                mv $$.ps ../$name.ps  
                gmt_cleanup .gmt  
                gmt_abort ~$0: First frame plotted to $name.ps~  
        fi  
        ps2raster $$.ps -Tt -E$dpi  
        mv $$.tif frames/$file.tif  
        echo ~Frame $file completed~  
        frame=‘gmt_set_framenext $frame‘  
done < $$.path.d  
if [ $# -eq 1 ]; then  
        echo ~anim_04.sh: Made $frame frames at 480x720 pixels placed in subdirectory frames~  
        convert $$/anim_0_123456.tiff ${name}_movie.m4v  
fi  
# 4. Clean up temporary files  
gmtset DOTS_PR_INCH 300  
gmt_cleanup .gmt
__________________________________________________________________________________________________

PIC

Figure 8.4: Flying over topography.

Chapter 9
Mailing lists, updates, and bug reports

Most public-domain (and even commercial) software comes with bugs, and the speed with which such bugs are detected and removed depends to a large degree on the willingness of the user community to report these to us in a useful manner. When your car breaks down, simply telling the mechanic that it doesn’t work will hardly speed up the repair or cut back costs! Therefore, we ask that if you detect a bug, first make sure that it in fact is a bug and not a user error. Then, post a new Issue about the problem. Be sure to include all the information necessary for us to recreate the situation in which the bug occurred. This will include the full command line used and, if possible, the data file used by the program. Post the bug-reports as a new issue on the GMT home page. We will try to fix bugs as soon as our schedules permit and inform users about the bug and availability of updated code (See Appendix D).

In addition to the bug tracker section we also maintain a user forum. It basically provides a place for GMT users to exchange ideas and ask questions about GMT usage, installation and portability, in other words all things other than plain bug reports (which should go to the Issues tracker). Please use this utility rather than sending questions directly to us personally. We hope you appreciate that we simply do not have time to be everybody’s personal GMT tutor.

While anyone can browse these pages, a registration is required for you to post to either the tracker or the forum. Register users will receive email about new GMT versions via the user forum, provided your notification settings are not modified to not receive such messages.

Appendix A
GMT supplemental packages

These packages are for the most part written and supported by us, but there are some exceptions. They provide extensions to GMT that are needed for particular rather than general applications. The packages are provided with GMT and are installed by default unless they requires non-GMT libraries; see the main GMT configuration process. Questions or bug reports for this software should be addressed to the person(s) listed in the README file associated with the particular program. It is not guaranteed that these programs are fully ANSI-C, Y2K, or POSIX compliant, or that they necessarily will install smoothly on all platforms, but most do. Note that the data sets some of these programs work on are not distributed with these packages; they must be obtained separately. The contents of the supplemental archive may change without notice; at this writing it contains these directories:

A.1 dbase: gridded data extractor

This package contains grdraster which you can use to extract data from global gridded data sets such as those available from NGDC. We have used it to prepare some of the grids in the examples (Chapter 6). You can also customize it to read your own data sets. The package is maintained by the GMT developers.

A.2 gshhg: GSHHG data extractor

This package contains gshhg which you can use to extract shoreline polygons from the Global Self-consistent Hierarchical High-resolution Geography (GSHHG) available separately from NGDC or the GSHHG home page (GSHHG contains both GSHHS and WDBII; GSHHS is the polygon data base from which the GMT coastlines derive while WDBII provides rivers and boundaries). It also contains gshhg_dp for cleverly decimating a shoreline, and gshhgtograss to convert shoreline segments to the GRASS database format; the latter program is maintained by Simon Cox32. The package is maintained by Paul Wessel.

A.3 imgsrc: gridded altimetry extractor

This package consists of the program img2mercgrd to extract subsets of the global gravity and predicted topography solutions derived from satellite altimetry33. The package is maintained by Walter Smith34.

A.4 meca: seismology and geodesy symbols

This package contains the programs pscoupe, psmeca, pspolar, and psvelo which are used by seismologists and geodesists for plotting focal mechanisms (including cross-sections and polarities), error ellipses, velocity arrows, rotational wedges, and more. The package is maintained by Kurt Feigl35 and Genevieve Patau36.

A.5 mex: Matlab/Octave–GMT interface

Here you will find the mex files grdinfo, grdread, and grdwrite, which can be used in Matlab or Octave to read and write grid files. The package originated with David Sandwell, UCSD, and was subsequently modified by Paul Wessel and Phil Sharfstein, UCSB. It is now maintained by Paul Wessel.

A.6 mgd77: MGD77 extractor and plotting tools

This package currently holds the programs mgd77convert, mgd77header, mgd77info, mgd77list, mgd77magref, mgd77manage, mgd77path, mgd77sniffer, and mgd77track which can be used to extract information or data values from or plot marine geophysical data files in the ASCII MGD77 or netCDF MGD77+ formats37). We expect this package eventually to replace the mgg package. The package is maintained by Paul Wessel.

A.7 mgg: GMT-MGD77 extractor and plotting tools

This package holds the legacy programs binlegs, dat2gmt, gmt2dat, gmtinfo, gmtlegs, gmtlist, gmtpath, gmttrack, and mgd77togmt, which can be used to maintain, access, extract data from, and plot marine geophysical data files converted from the MGD77 format to the .gmt format38). The package is maintained by the GMT developers.

A.8 misc: Miscellaneous tools

At the moment, this package contains the programs dimfilter, which is an extension of grdfilter in that it allows for spatial directional filtering, psmegaplot which you can use to make large posters using a simple laserwriter, makepattern which generates raster patterns from GMT 3.0 icon files, gmt2kml which converts GMT table data to Google Earth’s KML format, gmtdigitize which provides a GMT interface to a digitizing tablet via a serial port, gmtstitch which can be used to assemble pieces digitized lines into complete lines or polygons, gmtdp which performs line reduction using the Douglas-Peucker algorithm, kml2gmt which extracts GMT table data from Google Earth KML files, and nc2xy which can extract data from column-oriented netCDF files. The package is maintained by Paul Wessel. The increasingly popular utility ps2raster, which simplifies the rasterization of GMTPostScript to raster formats (see Appendix C), was moved to the general tools starting with GMT 4.2.0.

A.9 segyprogs: plotting SEGY seismic data

This package contains programs to plot SEGY seismic data files using the GMT mapping transformations and postscript library. pssegy generates a 2-D plot (x:location and y:time/depth) while pssegyz generates a 3-D plot (x and y: location coordinates, z: time/depth). Locations may be read from predefined or arbitrary portions of each trace header. Finally, segy2grd can convert SEGY data to a GMT grid file. The package is maintained by Tim Henstock39.

A.10 sph: spherical triangulation and gridding

This package contains the main programs sphtriangulate, which you can use to generate data for Delaunay or Voronoi diagrams, sphdistance which calculates distances from lines to grid nodes using Voronoi decomposition of the data, and sphinterpolate which performs gridding under tension on a sphere. These programs passes the heavy work onto the two Fortran-77 packages SSRFPACK and STRIPACK by Robert Renka; here they have been translated to C with assistance from f2c. The package is maintained by Paul Wessel.

A.11 spotter: backtracking and hotspotting

This package contains the plate tectonic programs backtracker, which you can use to move geologic markers forward or backward in time, grdrotater which rotates entire grids using a finite rotation, hotspotter which generates CVA grids based on seamount locations and a set of absolute plate motion stage poles (grdspotter does the same using a bathymetry grid instead of seamount locations), originator, which associates seamounts with the most likely hotspot origins, and rotconverter which does various operations involving finite rotations on a sphere. The package is maintained by Paul Wessel.

A.12 x2sys: track crossover error estimation

This package contains the tools x2sys_datalist, which allows you to extract data from almost any binary or ASCII data file, and x2sys_cross which determines crossover locations and errors generated by one or several geospatial tracks. Newly added are the tools x2sys_init, x2sys_binlist, x2sys_get, x2sys_list, x2sys_put, x2sys_report, x2sys_solve and x2sys_merge which extends the track-management system employed by the mgg supplement to generic track data of any format. This package represents a new generation of tools intended to replace the old “X_SYSTEM” crossover tools (below). The package is maintained by Paul Wessel.

A.13 x_system: track crossover error estimation

This package contains the tools x_edit, x_init, x_list, x_over, x_remove, x_report, x_setup, x_solve_dc_drift, and x_update. Collectively, they make up the old “XSYSTEM” crossover tools. This package with remain in the GMT supplemental archive until x2sys is complete. The package is maintained by Paul Wessel.

A.14 xgrid: visual editor for grid files

The package contains an X11 editor (xgridedit) for visual editing of grid files. It was originally developed by Hugh Fisher, CRES, in March 1992 but is now maintained by Lloyd Parkes40.

Appendix B
GMT file formats

B.1 Table data

These files have N records which have M fields each. Most programs can read multicolumn files, but require that the x [and y] variable(s) be stored in the 1st [and 2nd] column (There are, however, some exceptions to this rule, such as filter1d and sample1d). GMT can read both ASCII and binary table data.

B.1.1 ASCII tables

Optional file header records

The first data record may be preceded by 1 or more header records. When using such files, make sure to use the -H option and set the parameter N_HEADER_RECS in the .gmtdefaults4 file (System default is 1 header record if -H is set; you may also use -Hnrecs directly). Fields within a record must be separated by spaces, tabs, or commas. Each field can be an integer or floating-point number or a geographic coordinate string using the [+|-]dd[:mm[:ss]][W|S|N|E|w|s|n|e] format. Thus, 12:30:44.5W, 17.5S, 1:00:05, and 200:45E are all valid input strings.

Optional segment header records

When dealing with time- or (x,y)-series it is usually convenient to have each profile in separate files. However, this may sometimes prove impractical due to large numbers of profiles. An example is files of digitized lineations where the number of individual features may range into the thousands. One file per feature would in this case be unreasonable and furthermore clog up the directory. GMT provides a mechanism for keeping more than one profile in a file. Such files are called multiple segment files and are identical to the ones just outlined except that they have subheaders interspersed with data records that signal the start of a segment. The subheaders may be of any format, but all must have the same character in the first column. When using such files, you must specify the -m option. The unique character is by default ’ >’, but you can override that by appending your chosen character to the M option. E.g., -mH will look for subheaders starting with H, whereas -m’*’ will check for asterisks (The quotes are necessary since * has special meaning to UNIX). Some programs such as psxy will examine the subheaders to see if they contain -W and -G options for specifying pen and fill attributes for individual segments, -Z to change color via a cpt-file, or -L for label specifications. These settings (and occasionally others) will override the corresponding command line options.

B.1.2 Binary tables

GMT programs also support native binary tables to speed up input-output for i/o-intensive tasks like gridding and preprocessing. Files may have no header (hence the -H option cannot be used) and all data must either be single or double precision (no mixing allowed). Multiple segment files are allowed (-m) and the segment headers are assumed to be records where all the fields equal NaN. Flags appended to -m are ignored. The format and number of fields are specified with the -b option. Thus, for input you may set -bi[s][n], where s designates single precision (default is d for double) and n is the number of fields. For output, use -bo[s] (the programs know how many columns to write, unless you use -m in which case we need to know the number of output columns up front). If you need to swap the byte-order on either input or output you must use upper case S or D instead.

B.1.3 NetCDF tables

More and more programs are now producing binary data in the netCDF format, and so GMT programs started to support tabular netCDF data (files containing one or more 1-dimensional arrays) starting with GMT version 4.3.0. Because of the meta data contained in those files, reading them is much less complex than reading native binary tables, and even than ASCII tables. GMT programs will read as many 1-dimensional columns as are needed by the program, starting with the first 1-dimensional it can find in the file. To specifically specify which variables are to be read, append the suffix ?var1/var2/... to the netCDF file name or add the option -bicvar1/var2/..., where var1, var2, etc. are the names of the variables to be processed. The latter option is particularly practical when more than one file is read: the -bic option will apply to all files. Currently, GMT only reads, but does not write, netCDF tabular data.

B.2 Grid files

B.2.1 NetCDF files

By default, GMT stores 2-D grids as COARDS-compliant netCDF files. COARDS (which stands for Cooperative Ocean/Atmosphere Research Data Service) is a convention used by many agencies distributing gridded data for ocean and atmosphere research. Sticking to this convention allows GMT to read gridded data provided by other institutes and other programs. Conversely, other general domain programs will be able to read grids created by GMT. COARDS is a subset of a more extensive convention for netCDF data called CF-1.0 (Climate and Forecast, version 1.0). Hence, GMT grids are also automatically CF-1.0-compliant. However, since CF-1.0 has more general application than COARDS, not all CF-1.0 compliant netCDF files can be read by GMT.

The netCDF grid file in GMT has several attributes (See Table B.1) to describe the content. The routine that deals with netCDF grid files is sufficiently flexible so that grid files slightly deviating from the standards used by GMT can also be read.




Attribute
Description


Global attributes


Conventions COARDS, CF-1.0 (optional)


title Title (optional)


source How file was created (optional)


node_offset 0 for gridline node registration (default), 1 for pixel registration


x- and y-variable attributes


long_name Coordinate name (default: “Longitude” and “Latitude”)


units Unit of the coordinate (default: “degrees_east” and “degrees_north”)


actual_range Minimum and maximum x and y of region; if absent
(or valid_range) the first and last x- and y-values are queried


z-variable attributes


long_name Name of the variable (default: “z”)


units Unit of the variable (no default)


scale_factor Factor to multiply z with (default: 1)


add_offset Offset to add to scaled z (default: 0)


actual_range Minimum and maximum z (optional)


_FillValue Value associated with missing data points; if absent an
(or missing_value)appropriate default value is assumed, depending on data type.



Table B.1: Attributes of default GMT grid file in COARDS-compliant netCDF format.

By default, the first 2-dimensional variable in a netCDF file will by read as the z variable and the coordinate axes x and y will be determined from the dimensions of the z variable. GMT will recognize whether the y (latitude) variable increases or decreases. Both forms of data storage are handled appropriately.

For more information on the use of COARDS-compliant netCDF files, and on how to load multi-dimensional grids, read Section 4.18.

GMT also allows other formats to be read. In addition to the default netCDF format it can use binary floating points, short integers, bytes, and bits, as well as 8-bit Sun raster files (colormap ignored). Additional formats may be used by supplying read/write functions and linking these with the GMT libraries. The source file gmt_customio.c has the information that programmers will need to augment GMT to read custom grid files. We anticipate that the number of pre-programmed formats will increase as enterprising users implement what they need. See Section 4.17 for more information.

B.2.2 Gridline and Pixel node registration

Scanline format means that the data are stored in rows (y = constant) going from the “top” (y = ymax (north)) to the “bottom” (y = ymin (south)). Data within each row are ordered from “left” (x = xmin (west)) to “right” (x = xmax (east)). The node_offset signals how the nodes are laid out. The grid is always defined as the intersections of all x ( x = xmin,xmin+xinc,xmin+2xinc,,xmax ) and y ( y = ymin,ymin+yinc,ymin+2yinc,,ymax ) lines. The two scenarios differ in which area each data point represents. The default node registration in GMT is gridline node registration. Most programs can handle both types, and for some programs like grdimage a pixel registered file makes more sense. Utility programs like grdsample and grdproject will allow you to convert from one format to the other; grdedit can make changes to the grid header and convert a pixel- to a gridline-registred grid, or vice versa.

Gridline registration

In this registration, the nodes are centered on the grid line intersections and the data points represent the average value in a cell of dimensions (xincyinc) centered on the nodes (Figure B.1). In the case of grid line registration the number of nodes are related to region and grid spacing by

nx =(xmax-xmin)xinc+1 ny =(ymax-ymin)yinc+1

which for the example in Figure B.1 yields nx = ny = 4.


PIC

Figure B.1: Gridline registration of data nodes.

Pixel registration

Here, the nodes are centered in the grid cells, i.e., the areas between grid lines, and the data points represent the average values within each cell (Figure B.2. In the case of pixel registration the number of nodes are related to region and grid spacing by

nx =(xmax-xmin)xinc ny =(ymax-ymin)yinc

Thus, given the same region (-R), the pixel node registered grids have one less column and one less row than the grid line registered grids; here we find nx = ny = 3.


PIC

Figure B.2: Pixel registration of data nodes.

B.2.3 Boundary Conditions for operations on grids

GMT has the option to specify boundary conditions in some programs that operate on grids (grdsample -L; grdgradient -L; grdtrack -L; nearneighbor -L; grdview -L). The boundary conditions come into play when interpolating or computing derivatives near the limits of the region covered by the grid. The default boundary conditions used are those which are “natural” for the boundary of a minimum curvature interpolating surface. If the user knows that the data are periodic in x (and/or y), or that the data cover a sphere with x,y representing longitude,latitude, then there are better choices for the boundary conditions. Periodic conditions on x (and/or y) are chosen by specifying x (and/or y) as the boundary condition flags; global spherical cases are specified using the g (geographical) flag. Behavior of these conditions is as follows:

Periodic
conditions on x indicate that the data are periodic in the distance (xmax-xmin) and thus repeat values after every N = (xmax-xmin)xinc. Note that this implies that in a grid-registered file the values in the first and last columns are equal, since these are located at x = xmin and x = xmax, and there are N +1 columns in the file. This is not the case in a pixel-registered file, where there are only N and the first and last columns are located at xmin+xinc2 and xmax-xinc2. If y is periodic all the same holds for y.
Geographical
conditions indicate the following:

  1. If (xmax-xmin) 360 and also 180 modulo xinc = 0 then a periodic condition is used on x with a period of 360; else a default condition is used on the x boundaries.
  2. If condition 1 is true and also ymax = 90 then a “north pole condition” is used at ymax, else a default condition is used there.
  3. If condition 1 is true and also ymin = -90 then a “south pole condition” is used at ymin, else a default condition is used there.

“Pole conditions” use a 180° phase-shift of the data, requiring 180 modulo xinc = 0.

Default
boundary conditions are
2f = n2f = 0

on the boundary, where f(x,y) is represented by the values in the grid file, and n is the derivative in the direction normal to a boundary, and

2 = 2 x2 + 2 y2

is the two-dimensional Laplacian operator.

B.2.4 Native binary grid files

The old style native grid file format that was common in earlier version of GMT is still supported, although the use of netCDF files is strongly recommended. The file starts with a header of 892 bytes containing a number of attributes defining the content. The grdedit utility program will allow you to edit parts of the header of an existing grid file. The attributes listed in Table B.2 are contained within the header record in the order given (except the z-array which is not part of the header structure, but makes up the rest of the file). As this header was designed long before 64-bit architectures became available, the jump from the first three integers to the subsequent doubles in the structure does not occur on a 16-byte alignment. While GMT handles the reading of these structures correctly, enterprising programmers must take care to read this header correctly (see our code for details).




Parameter
Description


int nx Number of nodes in the x-dimension


int ny Number of nodes in the y-dimension


int node_offset 0 for grid line registration, 1 for pixel registration


double x_min Minimum x-value of region


double x_max Maximum x-value of region


double y_min Minimum y-value of region


double y_max Maximum y-value of region


double z_min Minimum z-value in data set


double z_max Maximum z-value in data set


double x_inc Node spacing in x-dimension


double y_inc Node spacing in y-dimension


double z_scale_factorFactor to multiply z-values after read


double z_add_offset Offset to add to scaled z-values


char x_units[80] Units of the x-dimension


char y_units[80] Units of the y-dimension


char z_units[80] Units of the z-dimension


char title[80] Descriptive title of the data set


char command[320] Command line that produced the grid file


char remark[160] Any additional comments




TYPE z[nx*ny] 1-D array with z-values in scanline format



Table B.2: GMT grid file header record. TYPE can be char, short, int, float, or double.

B.3 Sun raster files

The Sun raster file format consists of a header followed by a series of unsigned 1-byte integers that represents the bit-pattern. Bits are scanline oriented, and each row must contain an even number of bytes. The predefined 1-bit patterns in GMT have dimensions of 64 by 64, but other sizes will be accepted when using the -Gp|P option. The Sun header structure is outline in Table B.3.




Parameter
Description


int ras_magic Magic number


int ras_width Width (pixels) of image


int ras_height Height (pixels) of image


int ras_depth Depth (1, 8, 24, 32 bits) of pixel


int ras_length Length (bytes) of image


int ras_type Type of file; see RT_* below


int ras_maptype Type of colormap; see RMT_* below


int ras_maplengthLength (bytes) of following map



Table B.3: Structure of a Sun raster file.

After the header, the color map (if ras_maptype is not RMT_NONE) follows for ras_maplength bytes, followed by an image of ras_length bytes. Some related definitions are given in Table B.4.




Macro name
Description


RAS_MAGIC 0x59a66a95


RT_STANDARD 1 (Raw pixrect image in 68000 byte order)


RT_BYTE_ENCODED2 (Run-length compression of bytes)


RT_FORMAT_RGB 3 ([X]RGB instead of [X]BGR)


RMT_NONE 0 (ras_maplength is expected to be 0)


RMT_EQUAL_RGB 1 (red[ras_maplength/3],green[],blue[])



Table B.4: Sun macro definitions relevant to raster files.

Numerous public-domain programs exist, such as xv and convert (in the ImageMagick package), that will translate between various raster file formats such as tiff, gif, jpeg, and Sun raster. Raster patterns may be created with GMT plotting tools by generating PostScript plots that can be rasterized by ghostscript and translated into the right raster format.

Appendix C
Including GMT graphics into your documents

Now that you made some nice graphics with GMT, it is time to add them to a document, an article, a report, your dissertation, a poster, a web page, or a presentation. Of course, you could try the old-fashioned scissors and glue stick. More likely, you want to incorporate your graphics electronically into the document. Depending on the application, the GMT PostScript file will need to be converted to Encapsulated PostScript (EPS), Portable Document Format (PDF), or some raster format (e.g., JPEG, PNG, or TIFF) in order to incorporate them into the document.

A large number of questions to the GMT-Help mailing list are related to these rendering issues, showing that something as seemingly straightforward as incorporating a PostScript file into a document is a far from trivial exercise. This Chapter will show how to include GMT graphics into documents and how to achieve the best quality results.

C.1 Making GMT Encapsulated PostScript Files

GMT can produce both freeform PostScript files and the more restricted Encapsulated PostScript files (EPS). The former is intended to be sent to a printer or PostScript previewer, while the latter is intended to be included in another document (but should also be able to print and preview). You control what kind of PostScript that GMT produces by manipulating the PAPER_MEDIA parameter (see the gmtdefaults man page for how this is accomplished). Note that a freeform PostScript file may contain special operators (such as Setpagedevice) that is specific to printers (e.g., selection of paper tray). Some previewers (among them, Sun’s pageview) do not understand these valid instructions and may fail to image the file. Also, embedding freeform PostScript with such instructions in it into a larger document can create printing to fail. While you could choose another viewer (we recommend gv (ghostview)) to view single plots prepared by GMT, it is generally wiser anyhow to select EPS output when you are creating a plot intended for inclusion into a larger document. Some programs (and some publishers as well) do not allow the use of instructions like Setpagedevice as part of embedded graphics.

An EPS file that is to be placed into another document needs to have correct bounding box parameters. These are found in the PostScript Document Comment %%BoundingBox. Applications that generate EPS files should set these parameters correctly. Because GMT makes the PostScript files on the fly, often with several overlays, it is not possible to do so accurately. However, GMT does make an effort to ensure that the BoundingBox is large enough to contain the entire composite plot41. Therefore, if you need a “tight” BoundingBox you need to post-process your PostScript file. There are several ways in which this can be accomplished.

If you do not want to modify your illustration but just include it in a text document: many word processors (such as Microsoft Word, Corel WordPerfect, and Apple Pages) will let you include a PostScript file that you may place but not edit. Newer versions of those programs also allow you to include PDF versions of your graphics. Except for Pages, you will not be able to view the figure on-screen, but it will print correctly.

C.2 Converting GMT PostScript to PDF or raster images

Since Adobe’s PDF (Portable Document Format) seems to become the de facto standard for vector graphics, you are often well off converting GMT produced PostScript files to PDF. Being both vector formats (i.e., they basically describe all objects, text and graphics as lines and curves), such conversion sounds awfully straightforward and not worth a full section in this document. But experience has shown differently, since most converters cut corners by using the same tool (Aladdin’s ghostscript) with basic default options that are not devised to produce the best quality PDF files.

For some applications it is practical or even essential that you convert your PostScript file into a raster format, such as GIF (Graphics Interchange Format), TIFF (Tagged Image File Format), PNG (Portable Network Graphics), or JPEG (Joint Photographic Experts Group). A web page is better served with a raster image that will immediately show on a web browser, than with a PostScript file that needs to be downloaded to view, despite the better printing quality of the PostScript image. A less obvious reason to convert your image to a raster format is to by-pass PowerPoint’s rendering engine in case you want to embed the image into a presentation.

The are a number of programs that will convert PostScript files to PDF or raster formats, like Aladdin’s pstopdf, pbmplus’ pstoimg, or ImageMagick’s convert, most of which run ghostscript behind the scenes. The same is true for viewers like gv and Apple’s Preview. So a lot of the times when people report that their PostScript plot does not look right but prints fine, it is the way ghostscript is used with its most basic settings that is to blame.

C.2.1 When converting or viewing PostScript goes awry

Here are some notorious pitfalls with ghostscript (and other rendering programs for that matter).

Rendering.
When you are converting to a raster format, make sure you use a high enough resolution so that the pixels do not show when it is enlarged onto a screen or using a projector. The right choice of resolution depends on the application, but do not feel limited to the default 72 dpi (dots-per-inch) that is offered by most converters.
Image compression.
There are lossy and non-lossy compressions. A compression algorithm is called “lossy” when information is lost in the conversion: there is no way back to get the full original. The effect can be seen when there are sharp color transitions in your image: the edges will get blurry in order to allow a more efficient compression. JPEG uses a lossy compression, PNG is non-lossy, and TIFF generally does not use compression at all. We therefore recommend you convert to PNG if you need to rasterize your plot, and leave JPEG to photographs.
Embedded image compression.
When your GMT plot includes objects produced by grdimage, psimage or pslegend, they are seen as “images”. The default options of ghostscript will use a lossy compression (similar to JPEG) on those images when converting them to PDF objects. This can be avoided, however, by inhibiting the compression altogether, or using the non-lossy flate compression, similar to the one used in the old compress program. This compression is fully reversible, so that your image does not suffer any loss.
Auto-rotation.
The ghostscript engine has the annoying habit to automatically rotate an image produced with portrait orientation (using the -P option) so that the height is always larger than the width. So if you have an image that was printed in portrait mode but happens to have a width larger than height (for example a global map), it would suddenly get rotated. Again, this function needs to be switched off. Apple’s Preview uses the ghostscript engine and suffers from the same annoying habit. Oddly enough, ghostscript does not force landscape plots to be “horizontal”.
Anti-aliasing.
This is not something to worry about when converting to PDF, but certainly when producing raster images (discussed below). Anti-aliasing in this context means that the rendering tries to avoid aliasing, for example, sampling only the blacks in a black-and-white hachure. It does so by first oversampling the image and then using “gray-shades” when a target pixel is only partially white or black.

Clearly, this can lead to some unwanted results. First, all edges and lines get blurry and second, the assumption of a white background causes the gray shades to stand out when transferring the image to background with a different color (like the popular sleep-inducing blue in PowerPoint presentations). A more surprising effect of anti-aliasing is that the seams between tiles that make up the land mask when using pscoast will become visible. The anti-aliasing somehow decides to blur the edges of all polygons, even when they are seamlessly connected to other polygons.

It is therefore wise to overrule the default anti-aliasing option and over-sample the image yourself by choosing a higher resolution.

Including fonts.
When you are producing print-ready copy to publishers, they will often (and justifiably) ask that you include all fonts in your PDF document. Again, ghostscript (and all converters relying on that engine) will not do so by default.

C.2.2 Using ps2raster

The remedy to all the problems mentioned in the previous section is readily available to you in the form of the GMT utility ps2raster. It is designed to provide the best quality PDF and raster files using ghostscript as a rendering engine. The program ps2raster avoids anti-aliasing and lossy compression techniques that are default to ghostscript and includes the fonts into the resulting PDF file to ensure portability. By default the fonts are rendered at 720 dots-per-inch in a PDF file and images are sampled to 300 dpi, but that can be changed with the -E option. Simply run

ps2raster -A -P -Tf *.ps

to convert all PostScript files to PDF while cropping it to the smallest possible BoundingBox. Or use the -Tg option to convert your files to PNG.

The -P option of ps2raster may also come in handy. When you have not supplied the -P option in your first GMT plot command, your plot will be in Landscape mode. That means that the plot will be rotated 90 degrees (anti-clockwise) to fit on a Portrait mode page when coming out of the printer. The -P option of ps2raster will undo that rotation, so that you do not have to do so within your document. This will only affect Landscape plots; Portrait plots will not be rotated.

C.3 Examples

C.3.1 GMT graphics in LATEX

Nearly all illustrations in this GMT documentation were GMT-produced PostScript files. They were converted to PDF files using ps2raster and then included into a LATEX document that was processed with pdflatex to create the PDF document you are reading.

To add the graphics into the LATEX document we use the \includegraphics command supplied by the graphicx package. In the preamble of your LATEX document you will need to include the line

\usepackage{graphicx}

The inclusion of the graphics will probably be inside a floating figure environment; something like this

\begin{figure}  
   \includegraphics{myplot}  
   \caption{This is my first plot in \LaTeX.}  
   \label{fig:myplot}  
\end{figure}

Note that the \includegraphics command does not require you to add the suffix .pdf to the file name. If you run pdflatex, it will look automatically for myplot.pdf. If you run latex, it will use myplot.eps instead.

You can scale your plot using the options width=, height=, or scale=. In addition, if your original graphics was produced in Landscape mode (i.e., you did not use GMT’s -P option: not while plotting, nor in ps2raster), you will need to rotate the plot as well. For example,

\includegraphics[angle=-90,width=0.8\textwidth]{myplot}

will rotate the image 90° clockwise and scale it such that its width (after rotation) will be 80% of the width of the text column.

C.3.2 GMT graphics in PowerPoint


PIC

Figure C.1: Examples of rendered images in a PowerPoint presentation.


PIC

Figure C.2: PowerPoint’s “Format Picture” dialogue to set scale and rotation.

In Figure C.1 we have attempted to include Figure 7.20 into a PowerPoint presentation. First the PostScript file was converted to PDF (using ps2raster), then loaded into PowerPoint and the white background color was made transparent using the formatting toolbar (shown on the left side of Figure C.1). Clearly, when we let PowerPoint do the rendering, we do not get the best result:

  1. The anti-aliasing causes the tiles that make up the land to stand out. This is because the anti-aliasing algorithm blurs all edges, even when the tiles join seamlessly.
  2. The background color was assumed to be white, hence the text is “smoothed” using gray shades. Instead, shades of blue which would be appropriate for the background we are using.

On the central column of Figure C.1 we have included PNG versions of a portion of the same example. This shows the workings of anti-aliasing and different resolutions. All samples were obtained with convert. The one on the top uses all default settings, resulting in an anti-aliased image at 72 dpi resolution (very much like the PDF included directly into PowerPoint).

Just switching anti-aliasing off (middle) is clearly not an option either. It is true that we got rid of the gray blurring and the seams between the tiles, but without anti-aliasing the image becomes very blocky. The solution is to render the image at a higher resolution (e.g., 300 dpi) without anti-aliasing and then shrink the image to the appropriate size (bottom of the central column in Figure C.1). The scaling, rotation as well as the selection of the transparent color can be accomplished through the “Formatting” tool bar and the “Format Picture” dialogue box of PowerPoint (Figure C.2), which can be found by double clicking the included image (or selecting and right-clicking or control-clicking on a one-button mouse).

C.4 Concluding remarks

These examples do not constitute endorsements of the products mentioned above; they only represent our limited experience with adding PostScript to various types of documents. For other solutions and further help, please post messages to the user forum on GMT home page.

Appendix D
Availability of GMT and related code

D.1 Source distribution

All the source code, support data, PDF and HTML versions of all documentation (including UNIX manual pages) can be obtained by anonymous ftp from several mirror sites. We also maintain a GMT page on the World Wide Web (http://gmt.soest.hawaii.edu); see this page for installation directions which allow for a simplified, automatic install procedure (for the purchase of CD-R and DVD-R media, see http://www.geoware-online.com.)

The GMT compressed tar archives requires bzip2 to expand. If this utility is not installed on your system, you must obtained it by your system’s package manager or install it separately42. The GMT archives are as follows:

gmt-4.5.18.tar.bz2
Contains all GMT and supplemental source code needed for compilation, support files needed at run-time (cpt files, symbols and PostScript patterns), and all documentation (man pages, Cookbook and Technical Reference, and the tutorial), the data files used in the tutorial, and all the shell scripts and support data used in the Cookbook section.
gmt-4.5.18-non-gpl.tar.bz2
Contains Shewchuk’s triangle code which is not released under the GNU license.
gshhg-gmt-nc3-2.3.7.tar.bz2
Contains all resolutions (full, high, intermediate, low, and crude) of the GSHHG coastline, river, and border databases. Required to run GMT.

The netCDF library that makes up the backbone of the grid file i/o operations can be obtained from Unidata by downloading he file netcdf.tar.Z from the anonymous FTP directory of unidata.ucar.edu.

D.2 Pre-compiled Executables

For Windows users who just want executables we have three Windows installers available. Choose one of the first two and optionally the third:

gmt-4.5.18_install32.exe
The 32-bit install with all GMT executables (including supplements), the netCDF DLL, the example batch scripts and data, and all documentation in HTML format.
gmt-4.5.18_install64.exe
The 64-bit install with all GMT executables (including supplements), the netCDF DLL, the example batch scripts and data, and all documentation in HTML format.
gmt-4.5.18_pdf_install.exe
Installer for the optional GMT documentation in PDF format.
gshhg-2.3.7_install.exe
Installer with the complete set of GSHHG coastlines, rivers, and borders.

Usually, only one of the GMT 32- or 64-bit installers will be needed.

Appendix E
Predefined bit and hachure patterns in GMT

GMT provides 90 different bit and hachure patterns that can be selected with the -Gp or -GP option in most plotting programs. The left side of each image was created using -Gp, the right side shows the inverted version using -GP. These patterns are reproduced below at 300 dpi.

PIC

Appendix F
Chart of octal codes for characters

The characters and their octal codes in the Standard and ISOLatin1 encoded fonts are shown in Figure F.1. Dark gray areas signify codes reserved for control characters. In order to use all the extended characters (shown in the light gray boxes) you need to set CHAR_ENCODING to Standard+ or ISOLatin1+ in your .gmtdefaults4 file43.


PIC PIC

Figure F.1: Octal codes and corresponding symbols for StandardEncoding (left) and ISOLatin1Encoding (right) fonts.

The chart for the Symbol character set (GMT font number 12) and Pifont ZapfDingbats character set (font number 34) are presented in Figure F.2 below. The octal code is obtained by appending the column value to the \?? value, e.g., is \266 in the Symbol font. The euro currency symbol is \240 in the Symbol font and will print if your printer supports it (older printer’s firmware will not know about the euro).


PIC PIC

Figure F.2: Octal codes and corresponding symbols for Symbol (left) and ZapfDingbats (right) fonts.

Appendix G
PostScript fonts used by GMT

GMT uses the standard 35 fonts that come with most PostScript laserwriters. If your printer does not support some of these fonts, it will automatically substitute the default font (which is usually Courier). The following is a list of the GMT fonts:


PIC

Figure G.1: The standard 35 PostScript fonts recognized by GMT.

For the special fonts Symbol (12) and ZapfDingbats (34), see the octal charts in Appendix F. When specifying fonts in GMT, you can either give the entire font name or just the font number listed in this table. To change the fonts used in plotting basemap frames, see the man page for gmtdefaults. For direct plotting of text-strings, see the man page for pstext. To add additional fonts that you may have purchased or that are available at your institution, see instructions in the CUSTOM_font_info.d under the share/pslib directory.

Appendix H
Problems with display of GMT PostScript

GMT creates valid (so far as we know) Adobe PostScript Level 2. It does not use operators specific to Level 3 and should therefore produce output that should print on all PostScript printers44. Sometimes unexpected things happen when GMT output is sent to certain printers or displays. This section lists some things we have learned from experience, and some work-arounds. Note that many of these lessons are now rather old so hopefully these workarounds no longer apply to anybody...

H.1 PostScript driver bugs

When you try to display a PostScript file on a device, such as a printer or your screen, then a program called a PostScript device driver has to compute which device pixels should receive which colors (black or white in the case of a simple laser printer) in order to display the file. At this stage, certain device-dependent things may happen. These are not limitations of GMT or PostScript, but of the particular display device. The following bugs are known to us based on our experiences:

  1. Early versions of the Sun SPARCprinter software caused linewidth-dependent path displacement. We reported this bug and it has been fixed in newer versions of the software. Try using psxy to draw y = f(x) twice, once with a thin pen (-W1) and once with a fat pen (-W10); if they do not plot on top of each other, you have this kind of bug and need new software. The problem may also show up when you plot a mixture of solid and dashed (or dotted) lines of various pen thickness
  2. The first version of the HP Laserjet 4M (prior to Aug–93) had bugs in the driver program. The old one was PostScript SIMM, part number C2080-60001; the new one is called PostScript SIMM, part number C2080-60002. You need to get this one plugged into your printer if you have an HP LaserJet 4M.
  3. Apple Laserwriters with the older versions of Apple’s PostScript driver will give the error “limitcheck” and fail to plot when they encounter a path exceeding about 1000–1500 points. Try to get a newer driver from Apple, but if you can’t do that, set the parameter MAX_L1_PATH to 1000–1500 or even smaller in the file src/pslib_inc.h and recompile GMT. The number of points in a PostScript path can be arbitrarily large, in principle; GMT will only create paths longer than MAX_L1_PATH if the path represents a filled polygon or clipping path. Line-drawings (no fill) will be split so that no segment exceeds MAX_L1_PATH. This means psxy -G will issue a warning when you plot a polygon with more than MAX_L1_PATH points in it. It is then your responsibility to split the large polygon into several smaller segments. If pscoast gives such warnings and the file fails to plot you may have to select one of the lower resolution databases The path limitation exemplified by these Apple printers is what makes the higher-resolution coastlines for pscoast non-trivial: such coastlines have to be organized so that fill operations do not generate excessively large paths. Some HP PostScript cartridges for the Laserjet III also have trouble with paths exceeding 1500 points; they may successfully print the file, but it can take all night!
  4. 8-bit color screen displays (and programs which use only 8-bits, even on 24-bit monitors, such as Sun’s pageview under OpenWindows) may not dither cleverly, and so the color they show you may not resemble the color your PostScript file is asking for. Therefore, if you choose colors you like on the screen, you may be surprised to find that your plot looks different on the hardcopy printer or film writer. The only thing you can do is be aware of this, and make some test cases on your hardcopy devices and compare them with the screen, until you get used to this effect. (Each hardcopy device is also a little different, and so you will eventually find that you want to tune your color choices for each device.) The rgb color cube in example 11 may help.
  5. Some versions of Sun’s OpenWindows program pageview have only a limited number of colors available; the number can be increased somewhat by starting openwin with the option “openwin -cubesize large”.
  6. Finally, pageview seem to have problems understanding the setpagedevice operator. We recommend you only use pageview on EPS files or use gv instead.
  7. Many color hardcopy devices use CMYK color systems. GMT PostScript uses RGB (even if your cpt files are using HSV). The three coordinates of RGB space can be mapped into three coordinates in CMY space, and in theory K (black) is superfluous. But it is hard to get CMY inks to mix into a good black or gray, so these printers supply a black ink as well, hence CMYK. The PostScript driver for a CMYK printer should be smart enough to compute what portion of CMY can be drawn in K, and use K for this and remove it from CMY; however, some of them aren’t.
  8. In early releases of GMT we always used the PostScript command r g b setrgbcolor to specify colors, even if the color happened to be a shade of gray (r = g = b) or black (r = g = b = 0). One of our users found that black came out muddy brown when he used FreedomOfPress to make a Versatec plot of a GMT map. He found that if he used the PostScript command g setgray (where g is a graylevel) then the problem went away. Apparently, his installation of FreedomOfPress uses only CMY with the command setrgbcolor, and so 0 0 0 setrgbcolor tries to make black out of CMY instead of K. To fix this, in release 2.1 of GMT we changed some routines in pslib.c to check if (r = g and r = b), in which case g setgray is used instead of r g b setrgbcolor.
  9. Recent experience with some Tektronix Phaser printers and with commercial printing shops has shown that this substitution creates problems precisely opposite of the problems our Versatec user has. The Tektronix and commercial (we think it was a Scitex) machines do not use K when you say 0 setgray but they do when you say 0 0 0 setrgbcolor. We believe that these problems are likely to disappear as the various software developers make their codes more robust. Note that this is not a fault with GMT: r = g = b = 0 means black and should plot that way. Thus, the GMT source code as shipped to you checks whether r = g and r = b, in which case it uses setgray, else setrgbcolor. If your gray tones are not being drawn with K, you have two work-around options: (1) edit the source for pslib.c or (2) edit your PostScript file and try using setrgbcolor in all cases. The simplest way to do so is to redefine the setgray operator to use setrgbcolor. Insert the line

    /setgray {dup dup setrgbcolor} def

    immediately following the first line in the file (starts with %!PS.)

  10. Some color film writers are very sensitive to the brand of film. If black doesn’t look black on your color slides, try a different film.

H.2 Resolution and dots per inch

The parameter DOTS_PR_INCH can be set by the user through the .gmtdefaults4 file or gmtset. By default it is equal to the value in the gmt_defaults.h file, which is supplied with 300 when you get GMT from us. This seems a good size for most applications, but should ideally reflect the resolution of your hardcopy device (most laserwriters have at least 300 dpi, hence our default value). GMT computes what the plot should look like in double precision floating point coordinates, and then converts these to integer coordinates at DOTS_PR_INCH resolution. This helps us find out that certain points in a path lie on top of other points, and we can remove these, making smaller paths. Small paths are important for the laserwriter bugs above, and also to make fill operations compute faster. Some users have set their DOTS_PR_INCH to very large numbers. This only makes the PostScript output bigger without affecting the appearance of the plot. However, if you want to make a plot which fits on a page at first, and then later magnify this same PostScript file to a huge size, the higher DPI is important. Your data may not have the higher resolution but on certain devices the edges of fonts will not look crisp if they are not drawn with an effective resolution of 300 dpi or so. Beware of making an excessively large path. Note that if you change dpi the linewidths produced by your -W options will change, unless you have appended p for linewidth in points.

H.3 European characters

Note for users of pageview in Sun OpenWindows: GMT now offers some octal escape sequences to load European alphabet characters in text strings (see Section 4.16). When this feature is enabled, the header on GMT PostScript output includes a section defining special fonts. The definition is added to the header whether or not your plot actually uses the fonts.

Users who view their GMT PostScript output using pageview in OpenWindows on Sun computers or user older laserwriters may have difficulties with the European font definition. If your installation of OpenWindows followed a space-saving suggestion of Sun, you may have excluded the European fonts, in which case pageview will fail to render your plot.

Ask your system administrator about this, or run this simple test: (1) View a GMT PostScript file with pageview. If it comes up OK, you will be fine. If it comes up blank, open the “Edit PostScript” button and examine the lower window for error messages. (The European font problem generates lots of error messages in this window). (2) Verify that the PostScript file is OK, by sending it to a laserwriter and making sure it comes out. (3) If the PostScript file is OK but it chokes pageview, then edit the PostScript file, cutting out everything between the lines:

%%%%% START OF EUROPEAN FONT DEFINITION %%%%%
<bunch of definitions >
%%%%% END OF EUROPEAN FONT DEFINITION %%%%%

Now try pageview on the edited version. If it now comes up, you have a limited subset of OpenWindows installed. If you discover that these fonts cause you trouble, then you can edit your .gmtdefaults4 file to set CHAR_ENCODING = Standard, which will suppress the printing of this definition in the GMT PostScript header. You can make output which will be viewable in pageview without any editing. However, you would have to reset this to TRUE before attempting to use European fonts, and then the output will become un-pageview-able again. If you try to concatenate segments of GMT PostScript made with and without the European fonts enabled, then you may find that you have problems, either with the definition, or because you ask for something not defined.

H.4 Hints

When making images and perspective views of large amounts of data, the GMT programs can take some time to run, the resulting PostScript files can be very large, and the time to display the plot can be long. Fine tuning a plot script can take lots of trial and error. We recommend using grdsample to make a low resolution version of the data files you are plotting, and practice with that, so it is faster; when the script is perfect, use the full-resolution data files. We often begin building a script using only psbasemap or pscoast to get the various plots oriented correctly on the page; once this works we replace the psbasemap calls with the actually desired GMT programs.

If you want to make color shaded relief images and you haven’t had much experience with it, here is a good first cut at the problem: Set your COLOR_MODEL to HSV using gmtset. Use makecpt or grd2cpt to make a continuous color palette spanning the range of your data. Use the -Nt option on grdgradient. Try the result, and then play with the tuning of the .gmtdefaults4, the cpt file, and the gradient file.

Appendix I
Color Space: The final frontier

In this Appendix, we are going to try to explain the relationship between the RGB, CMYK, and HSV color systems so as to (hopefully) make them more intuitive. GMT allows users to specify colors in cpt files in either of these three systems. Interpolation between colors is performed in either RGB or HSV, depending on the specification in the cpt file. Below, we will explain why this all matters.

I.1 RGB color system

Remember your (parents’) first color television set? Likely it had three little bright colored squares on it: red, green, and blue. And that is exactly what each color on the tube is made of: varying levels of red, green and blue light. Switch all of them off, r = g = b = 0, then you have black. All of them at maximum, r = g = b = 255, creates white. Your computer screen works the same way.

A mix of levels of red, green, and blue creates basically any color imaginable. In GMT each color can be represented by the triplet r/g/b. For example, 127/255/0 (half red, full green, and no blue) creates a color called chartreuse. The color sliders in the graphics program GIMP are an excellent way to experiment with colors, since they show you in advance how moving one of the color sliders will change the color. As Figure I.1a shows: increase the red and you will get a more yellow color, while lowering the blue level will turn it into brown.


PIC a b PIC

Figure I.1: Chartreuse in GIMP. (a) Sliders indicate how the color is altered when changing the H, S, V, R, G, or B levels. (b) For a constant hue (here 90°) value increases to the right and saturation increases up, so the “pure” color is on the top right.

Is chocolate your favorite color, but you do not know the RGB equivalent values? Then look them up in Figure I.2 or type man gmtcolors for a full list. It’s 210/105/30. But GMT makes it easy on you: you can specify pen, fill, and palette colors by any of the more than 500 unique colors found in that file.

Are you very web-savvy and work best with hexadecimal color codes as they are used in HTML? Even that is allowed in GMT. Just start with a hash mark (#) and follow with the 2 hexadecimal characters for red, green, and blue. For example, you can use #79ff00 for chartreuse, #D2691E for chocolate.


PIC

Figure I.2: The 555 unique color names that can be used in GMT. Lower, upper, or mixed case, as well as the british spelling of “grey” are allowed. A4, Letter, and Tabloid sized versions of this RGB chart can be found in the GMT documentation directory.

I.2 HSV color system

If you have played around with RGB color sliders, you will have noticed that it is not intuitive to make a chosen color lighter or darker, more saturated or more gray. It would involve changing three sliders. To make it easier to manipulate colors in terms of lightness and saturation, another coordinate system was invented: HSV (hue, saturation, value). Those terms can be made clear best by looking at the color sliders in Figure I.1a. Hue (running from 0° to 360°) gives you the full spectrum of saturated colors. Saturation (from 0 to 1, or 100%) tells you how ‘full’ your color is: reduce it to zero and you only have gray scales. Value (from 0 to 1, or 100%) will bring you from black to a fully saturated color. Note that “value” is not the same as “intensity”, or “lightness”, used in other color geometries. “Brilliance” may be the best alternative word to describe “value”. Apple calls it as “brightness”, and hence refers to HSB for this color space.

Want more chartreuse or chocolate? You can specify them in GMT as 90-1-1 and 25-0.86-0.82, respectively.

I.3 The color cube

We are going to try to give you a geometric picture of color mixing in RGB and HSV by means of a tour of the RGB cube depicted in Figure 7.11. The geometric picture is most helpful, we think, since HSV are not orthogonal coordinates and not found from RGB by a simple algebraic transformation. So here goes: Look at the cube face with black, red, magenta, and blue corners. This is the g = 0 face. Orient the cube so that you are looking at this face with black in the lower left corner. Now imagine a right-handed cartesian (r,g,b) coordinate system with origin at the black point; you are looking at the g = 0 plane with r increasing to your right, g increasing away from you, and b increasing up. Keep this sense of (r,g,b) as you look at the cube.

Now tip the cube such that the black corner faces down and the white corner up. When looking from the top, you can see the hue, contoured in gray solid lines, running around in 360° counter-clockwise. It starts with shades of red (0°), then goes through green (120°) and blue (240°), back to red.

On the three faces that are now on the lower side (with the white print) one of (r,g,b) is equal to 0. These three faces meet at the black corner, where r = g = b = 0. On these three faces the colors are fully saturated: s = 1. The dashed white lines indicate different levels of v, ranging from 0 to 1 with contours every 0.1.

On the upper three faces (with the black print), one of (r,g,b) is equal to the maximum value. These three faces meet at the white corner, where r = g = b = 255. On these three faces value is at its maximum: v = 1 (or 100%). The dashed black lines indicate varying levels of saturation: s ranges from 0 to 1 with contours every 0.1.

Now turn the cube around on its vertical axis (running from the black to the white corner). Along the six edges that zigzag around the “equator”, both saturation and value are maximum, so s = v = 1. Twirling the cube around and tracing the zigzag, you will visit six of the eight corners of the cube, with changing hue (h): red (0°), yellow (60°), green (120°), cyan (180°), blue (240°), and magenta (300°). Three of these are the RGB colors; the other three are the CMY colors which are the complement of RGB and are used in many color hardcopy devices (see below). The only cube corners you did not visit on this path are the black and white corners. They lie on the vertical axis where hue is undefined and r = g = b. Any point on this axis is a shade of gray.

Let us call the points where s = v = 1 (points along the RYGCBM path described above) the “pure” colors. If we start at a pure color and we want to whiten it, we can keep h constant and v = 1 while decreasing s; this will move us along one of the cube faces toward the white point. If we start at a pure color and we want to blacken it, we can keep h constant and s = 1 while decreasing v; this will move us along one of the cube faces toward the black point. Any point in (r,g,b) space which can be thought of as a mixture of pure color + white, or pure color + black, is on a face of the cube.

The points in the interior of the cube are a little harder to describe. The definition for h above works at all points in (non-gray) (r,g,b) space, but so far we have only looked at (s, v) on the cube faces, not inside it. At interior points, none of (r,g,b) is equal to either 0 or 255. Choose such a point, not on the gray axis. Now draw a line through your point so that the line intersects the gray axis and also intersects the RYGCBM path of edges somewhere. It is always possible to construct this line, and all points on this line have the same hue. This construction shows that any point in RGB space can be thought of as a mixture of a pure color plus a shade of gray. If we move along this line away from the gray axis toward the pure color, we are “purifying” the color by “removing gray”; this move increases the color’s saturation. When we get to the point where we cannot remove any more gray, at least one of (r,g,b) will have become zero and the color is now fully saturated; s = 1. Conversely, any point on the gray axis is completely undersaturated, so that s = 0 there. Now we see that the black point is special, s is both 0 and 1 at the same time. In other words, at the black point saturation in undefined (and so is hue). The convention is to use h = s = v = 0 at this point.

It remains to define value. To do so, try this: Take your point in RGB space and construct a line through it so that this line goes through the black point; produce this line from black past your point until it hits a face on which v = 1. All points on this line have the same hue. Note that this line and the line we made in the previous paragraph are both contained in the plane whose hue is constant. These two lines meet at some arbitrary angle which varies depending on which point you chose. Thus HSV is not an orthogonal coordinate system. If the line you made in the previous paragraph happened to touch the gray axis at the black point, then these two lines are the same line, which is why the black point is special. Now, the line we made in this paragraph illustrates the following: If your chosen point is not already at the end of the line, where v = 1, then it is possible to move along the line in that direction so as to increase (r,g,b) while keeping the same hue. The effect this has on a color monitor is to make the color more “brilliant”, your hue will become “stronger”; if you are already on a plane where at least one of (r,g,b) = 255, then you cannot get a stronger version of the same hue. Thus, v measures brilliance or strength. Note that it is not quite true to say that v measures distance away from the black point, because v is not equal to r2 +g2 +b2255.

Another representation of the HSV space is the color cone illustrated in Figure I.3.


“Pure” colors are around the edge of the circular surface at the top. Hue runs counter-clockwise. Saturation decreases to the center. Value increases from zero (black) at the bottom to 1 at the top. Gray shades are along the vertical axis. PIC

Figure I.3: The HSV color space.

I.4 Color interpolation

From studying the RGB cube, we hope you will have understood that there are different routes to follow between two colors, depending whether you are in the RGB or HSV system. Suppose you would make an interpolation between blue and red. In the RGB system you would follow a path diagonally across a face of the cube, from 0/0/255 (blue) via 127/0/127 (purple) to 255/0/0 (red). In the HSV system, you would trace two edges, from 240-1-1 (blue) via 300-1-1 (magenta) to 360-1-1 (red). That is even assuming software would be smart enough to go the shorter route. More likely, red will be recorded as 0-1-1, so hue will be interpolated the other way around, reducing hue from 240° to 0°, via cyan, green, and yellow.

Depending on the design of your color palette, you may want to have it either way. By default, GMT interpolates in RGB space, even when the original color palette is in the HSV system. However, when you add the line #COLOR_MODEL=+HSV (with the leading ‘+’ sign) in the header of the color palette file, GMT will not only read the color representation as HSV values, but also interpolate colors in the HSV system. That means that H, S, and V values are interpolated linearly between two colors, instead of their respective R, G, and B values.

The top row in Figure I.4 illustrates two examples: a blue-white-red scale (the polar palette in Appendix M) interpolated in RGB and the rainbow palette interpolated in HSV. The bottom row of the Figure demonstrates how things can go terribly wrong when you do the interpolation in the other system.


PIC

Figure I.4: When interpolating colors, the color system matters. The polar palette on the left needs to be interpolated in RGB, otherwise hue will change between blue (240°) and white (0°). The rainbow palette should be interpolated in HSV, since only hue should change between magenta (300°) and red (0°). Diamonds indicate which colors are defined in the palettes; they are fixed, the rest is interpolated.

I.5 Artificial illumination

GMT uses the HSV system to achieve artificial illumination of colored images (e.g., -I option in grdimage) by changing the saturation s and value v coordinates of the color. When the intensity is zero (flat illumination), the data are colored according to the cpt file. If the intensity is non-zero, the color is either lightened or darkened depending on the illumination. The color is first converted to HSV (if necessary) and then darkened by moving (s,v) toward (HSV_MIN_SATURATION, HSV_MIN_VALUE) if the intensity is negative, or lightened by sliding (s,v) toward (HSV_MAX_SATURATION, HSV_MAX_VALUE) if the illumination is positive. The extremes of the s and v are defined in the .gmtdefaults4 file and are usually chosen so the corresponding points are nearly black (s = 1, v = 0) and white (s = 0, v = 1). The reason this works is that the HSV system allows movements in color space which correspond more closely to what we mean by “tint” and “shade”; an instruction like “add white” is easy in HSV and not so obvious in RGB.

I.6 Thinking in RGB or HSV

The RGB system is understandable because it is cartesian, and we all learned cartesian coordinates in school. But it doesn’t help us create a tint or shade of a color; we cannot say, “We want orange, and a lighter shade of orange, or a less vivid orange”. With HSV we can do this, by saying, “Orange must be between red and yellow, so its hue is about h = 30°; a less vivid orange has a lesser s, a darker orange has a lesser v”. On the other hand, the HSV system is a peculiar geometric construction, more like a cone (Figure I.3). It is not an orthogonal coordinate system, and it is not found by a matrix transformation of RGB; these make it difficult in some cases too. Note that a move toward black or a move toward white will change both s and v, in the general case of an interior point in the cube. The HSV system also doesn’t behave well for very dark colors, where the gray point is near black and the two lines we constructed above are almost parallel. If you are trying to create nice colors for drawing chocolates, for example, you may be better off guessing in RGB coordinates.

I.7 CMYK color system

Finally, you can imagine that printers work in a different way: they mix different paints to make a color. The more paint, the darker the color, which is the reverse of adding more light. Also, mixing more colored paints does not give you true black, so that means that you really need four colors to do it right. Open up your color printer and you’ll probably find four cartridges: cyan, magenta, yellow (often these are combined into one), and black. They form the CMYK system of colors, each value running from 0 to 1 (or 100%). In GMT CMYK color coding can be achieved using c/m/y/k quadruplets.

Obviously, there is no unique way to go from the 3-dimensional RGB system to the 4-dimensional CMYK system. So, again, there is a lot of hand waving applied in the transformation. Strikingly, CMYK actually covers a smaller color space than RGB. We will not try to explain you the details behind it, just know that there is a transformation needed to go from the colors on your screen to the colors on your printer. It might explain why what you see is not necessarily what you get. If you are really concerned about how your color plots will show up in your PhD thesis, for example, it might be worth trying to save and print all your color plots using the CMYK system. Letting GMT do the conversion to CMYK may avoid some nasty surprises when it comes down to printing. To specify the color space of your PostScript file, set PS_COLOR in the .gmtdefaults4 file to RGB, HSV, or CMYK.

Appendix J
Filtering of data in GMT

The GMT programs filter1d (for tables of data indexed to one independent variable) and grdfilter (for data given as 2-dimensional grids) allow filtering of data by a moving-window process. (To filter a grid by Fourier transform use grdfft.) Both programs use an argument -F <type ><width > to specify the type of process and the window’s width (in 1-d) or diameter (in 2-d). (In filter1d the width is a length of the time or space ordinate axis, while in grdfilter it is the diameter of a circular area whose distance unit is related to the grid mesh via the -D option). If the process is a median, mode, or extreme value estimator then the window output cannot be written as a convolution and the filtering operation is not a linear operator. If the process is a weighted average, as in the boxcar, cosine, and gaussian filter types, then linear operator theory applies to the filtering process. These three filters can be described as convolutions with an impulse response function, and their transfer functions can be used to describe how they alter components in the input as a function of wavelength.

Impulse responses are shown here for the boxcar, cosine, and gaussian filters. Only the relative amplitudes of the filter weights shown; the values in the center of the window have been fixed equal to 1 for ease of plotting. In this way the same graph can serve to illustrate both the 1-d and 2-d impulse responses; in the 2-d case this plot is a diametrical cross-section through the filter weights (Figure J.1).


PIC

Figure J.1: Impulse responses for GMT filters.

Although the impulse responses look the same in 1-d and 2-d, this is not true of the transfer functions; in 1-d the transfer function is the Fourier transform of the impulse response, while in 2-d it is the Hankel transform of the impulse response. These are shown in Figures J.2 and J.3, respectively. Note that in 1-d the boxcar transfer function has its first zero crossing at f = 1, while in 2-d it is around f ~ 1.2. The 1-d cosine transfer function has its first zero crossing at f = 2; so a cosine filter needs to be twice as wide as a boxcar filter in order to zero the same lowest frequency. As a general rule, the cosine and gaussian filters are “better” in the sense that they do not have the “side lobes” (large-amplitude oscillations in the transfer function) that the boxcar filter has. However, they are correspondingly “worse” in the sense that they require more work (doubling the width to achieve the same cut-off wavelength).


PIC

Figure J.2: Transfer functions for 1-D GMT filters.

One of the nice things about the gaussian filter is that its transfer functions are the same in 1-d and 2-d. Another nice property is that it has no negative side lobes. There are many definitions of the gaussian filter in the literature (see page 7 of Bracewell45). We define σ equal to 1/6 of the filter width, and the impulse response proportional to exp[-0.5(tσ)2). With this definition, the transfer function is exp[-2(πσf)2] and the wavelength at which the transfer function equals 0.5 is about 5.34 σ, or about 0.89 of the filter width.


PIC

Figure J.3: Transfer functions for 2-D (radial) GMT filters.

Appendix K
The GMT High-Resolution Coastline Data

Starting with version 3.0, GMT use a completely new coastline database and the pscoast utility was been completely rewritten to handle the new file format. Many users have asked us why it has taken so long for GMT to use a high-resolution coastline database; after all, such data have been available in the public domain for years. To answer such questions we will take you along the road that starts with these public domain data sets and ends up with the database used by GMT.

K.1 Selecting the right data

There are two well-known public-domain data sets that could be used for this purpose. Once is known as the World Data Bank II or CIA Data Bank (WDB) and contains coastlines, lakes, political boundaries, and rivers. The other, the World Vector Shoreline (WVS) only contains shorelines between saltwater and land (i.e., no lakes). It turns out that the WVS data is far superior to the WDB data as far as data quality goes, but as noted it lacks lakes, not to mention rivers and borders. We decided to use the WVS whenever possible and supplement it with WDB data. We got these data over the Internet; they are also available on CD-ROM from the National Geophysical Data Center in Boulder, Colorado46.

K.2 Format required by GMT

In order to paint continents or oceans it is necessary that the coastline data be organized in polygons that may be filled. Simple line segments can be used to draw the coastline, but for painting polygons are required. Both the WVS and WDB data consists of unsorted line segments: there is no information included that tells you which segments belong to the same polygon (e.g., Australia should be one large polygon). In addition, polygons enclosing land must be differentiated from polygons enclosing lakes since they will need different paint. Finally, we want pscoast to be flexible enough that it can paint the land or the oceans or both. If just land (or oceans) is selected we do not want to paint those areas that are not land (or oceans) since previous plot programs may have drawn in those areas. Thus, we will need to combine polygons into new polygons that lend themselves to fill land (or oceans) only (Note that older versions of pscoast always painted lakes and wiped out whatever was plotted beneath).

K.3 The long and winding road

The WVS and WDB together represent more than 100 Mb of binary data and something like 20 million data points. Hence, it becomes obvious that any manipulation of these data must be automated. For instance, the reasonable requirement that no coastline should cross another coastline becomes a complicated processing step.

  1. To begin, we first made sure that all data were “clean”, i.e., that there were no outliers and bad points. We had to write several programs to ensure data consistency and remove “spikes” and bad points from the raw data. Also, crossing segments were automatically “trimmed” provided only a few points had to be deleted. A few hundred more complicated cases had to be examined semi-manually.
  2. Programs were written to examine all the loose segments and determine which segments should be joined to produce polygons. Because not all segments joined exactly (there were non-zero gaps between some segments) we had to find all possible combinations and choose the simplest combinations. The WVS segments joined to produce more than 200,000 polygons, the largest being the Africa-Eurasia polygon which has 1.4 million points. The WDB data resulted in a smaller data base ( ~25% of WVS).
  3. We now needed to combine the WVS and WDB data bases. The main problem here is that we have duplicates of polygons: most of the features in WVS are also in WDB. However, because the resolution of the data differ it is nontrivial to figure out which polygons in WDB to include and which ones to ignore. We used two techniques to address this problem. First, we looked for crossovers between all possible pairs of polygons. Because of the crossover processing in step 1 above we know that there are no remaining crossovers within WVS and WDB; thus any crossovers would be between WVS and WDB polygons. Crossovers could mean two things: (1) A slightly misplaced WDB polygon crosses a more accurate WVS polygon, both representing the same geographic feature, or (2) a misplaced WDB polygon (e.g., a small coastal lake) crosses the accurate WVS shoreline. We distinguished between these cases by comparing the area and centroid of the two polygons. In almost all cases it was obvious when we had duplicates; a few cases had to be checked manually. Second, on many occasions the WDB duplicate polygon did not cross its WVS counterpart but was either entirely inside or outside the WVS polygon. In those cases we relied on the area-centroid tests.
  4. While the largest polygons were easy to identify by visual inspection, the majority remain unidentified. Since it is important to know whether a polygon is a continent or a small pond inside an island inside a lake we wrote programs that would determine the hierarchical level of each polygon. Here, level = 1 represents ocean/land boundaries, 2 is land/lakes borders, 3 is lakes/islands-in-lakes, and 4 is islands-in-lakes/ponds-in-islands-in-lakes. Level 4 was the highest level encountered in the data. To automatically determine the hierarchical levels we wrote programs that would compare all possible pairs of polygons and find how many polygons a given polygon was inside. Because of the size and number of the polygons such programs would typically run for 3 days on a Sparc-2 workstation.
  5. Once we know what type a polygon is we can enforce a common “orientation” for all polygons. We arranged them so that when you move along a polygon from beginning to end, your left hand is pointing toward “land”. At this step we also computed the area of all polygons since we would like the option to plot only features that are bigger than a minimum area to be specified by the user.
  6. Obviously, if you need to make a map of Denmark then you do not want to read the entire 1.4 million points making up the Africa-Eurasia polygon. Furthermore, most plotting devices will not let you paint and fill a polygon of that size due to memory restrictions. Hence, we need to partition the polygons so that smaller subsets can be accessed rapidly. Likewise, if you want to plot a world map on a letter-size paper there is no need to plot 10 million data points as most of them will plot several times on the same pixel and the operation would take a very long time to complete. We chose to make 5 versions on the database, corresponding to different resolutions. The decimation was carried out using the Douglas-Peucker (DP) line-reduction algorithm47. We chose the cutoffs so that each subset was approximately 20% the size of the next higher resolution. The five resolutions are called full, high, intermediate, low, and crude; they are accessed in pscoast, gmtselect, and grdlandmask with the -D option48. For each of these 5 data sets (f, h, i, l, c) we specified an equidistant grid (1°, 2°, 5°, 10°, 20°) and split all polygons into line-segments that each fit inside one of the many boxes defined by these grid lines. Thus, to paint the entire continent of Australia we instead paint many smaller polygons made up of these line segments and gridlines. Some book-keeping has to be done since we need to know which parent polygon these smaller pieces came from in order to prescribe the correct paint or ignore if the feature is smaller than the cutoff specified by the user. The resulting segment coordinates were then scaled to fit in short integer format to preserve precision and written in netCDF format for ultimate portability across hardware platforms49.
  7. While we are now back to a file of line-segments we are in a much better position to create smaller polygons for painting. Two problems must be overcome to correctly paint an area:

    GMT uses a polygon-assembly routine that carries out these tasks on the fly.

K.4 The Five Resolutions

We will demonstrate the power of the new database by starting with a regional hemisphere map centered near Papua New Guinea and zoom in on a specified point. The map regions will be specified in projected km from the projection center, e.g., we may want the map to go from -2000 km to +2000 km in the longitudinal and the latitudinal direction. However, GMT programs expects degrees in the -R option that specifies the desired region. Given the chosen map projection we can automate this process by using a simple shell function that we call getbox:

_________________________________________________________________________________

getbox () {  
# Expects -Joption and distance in km from map center  
range=‘(echo -$2 -$2; echo $2 $2) | mapproject $1 -R0/360/-90/90 -I -Fk -C‘  
printf ~ -R%f/%f/%f/%fr\n~ $range  
}
__________________________________________________________________________________________________

Also, as we zoom in on the projection center we want to draw the outline of the next map region on the plot. To do that we need the geographical coordinates of the four corners of the region rectangle. Again, we automate this task by using our simple function getrect:

_________________________________________________________________________________

getrect () {  
# Expects xmin xmax ymin ymax in km relative to map center  
# -R and -J are set by preceding GMT commands  
(echo -$1 -$1; echo -$1 $1; echo $1 $1; echo $1 -$1) | mapproject -R -J -I -Fk -C  
}
__________________________________________________________________________________________________

K.4.1 The crude resolution (-Dc)

We begin with an azimuthal equidistant map of the hemisphere centered on 130°21’E, 0°12’S, which is slightly west of New Guinea, near the Strait of Dampier. The edges of the map are all 9000 km true distance from the projection center. At this scale (and for global maps) the crude resolution data will usually be adequate to capture the main geographic features. To avoid cluttering the map with insignificant detail we only plot features (i.e., polygons) that exceed 500 km2 in area. Smaller features would only occupy a few pixels on the plot and make the map look “dirty”. We also add national borders to the plot. The crude database is heavily decimated and simplified by the DP-routine: The total file size of the coastlines, rivers, and borders database is only 283 kbytes. The plot is produced by the script:

_________________________________________________________________________________

gmtset GRID_CROSS_SIZE_PRIMARY 0 OBLIQUE_ANNOTATION 22 ANNOT_MIN_SPACING 0.3  
pscoast ‘getbox -JE130.35/-0.2/3.5i 9000‘ -J -P -Dc \  
        -A500 -Glightgray -Wthinnest -N1/thinnest,- -B20g20WSne -K > GMT_App_K_1.ps  
getrect 2000 | psxy -R -J -O -Wthicker -L -A >> GMT_App_K_1.ps
__________________________________________________________________________________________________

PIC

Figure K.1: Map using the crude resolution coastline data.

Here, we use the OBLIQUE_ANNOTATION bit flags to achieve horizontal annotations and set ANNOT_MIN_SPACING to suppress some longitudinal annotations near the S pole that otherwise would overprint. The box indicates the outline of the next map.

K.4.2 The low resolution (-Dl)

We have now reduced the map area by zooming in on the map center. Now, the edges of the map are all 2000 km true distance from the projection center. At this scale we choose the low resolution data that faithfully reproduce the dominant geographic features in the region. We cut back on minor features less than 100 km2 in area. We still add national borders to the plot. The low database is less decimated and simplified by the DP-routine: The total file size of the coastlines, rivers, and borders combined grows to 907 kbytes; it is the default resolution in GMT. The plot is generated by the script:

_________________________________________________________________________________

pscoast ‘getbox -JE130.35/-0.2/3.5i 2000‘ -J -P -Dl -A100 \  
        -Glightgray -Wthinnest -N1/thinnest,- -B10g5WSne -K > GMT_App_K_2.ps  
getrect 500 | psxy -R -J -O -Wthicker -L -A >> GMT_App_K_2.ps
__________________________________________________________________________________________________

PIC

Figure K.2: Map using the low resolution coastline data.

K.4.3 The intermediate resolution (-Di)

We continue to zoom in on the map center. In this map, the edges of the map are all 500 km true distance from the projection center. We abandon the low resolution data set as it would look too jagged at this scale and instead employ the intermediate resolution data that faithfully reproduce the dominant geographic features in the region. This time, we ignore features less than 20 km2 in area. Although the script still asks for national borders none exist within our region. The intermediate database is moderately decimated and simplified by the DP-routine: The combined file size of the coastlines, rivers, and borders now exceeds 3.35 Mbytes. The plot is generated by the script:

_________________________________________________________________________________

pscoast ‘getbox -JE130.35/-0.2/3.5i 500‘ -J -P -Di -A20 \  
        -Glightgray -Wthinnest -N1/thinnest,- -B2g1WSne -K > GMT_App_K_3.ps  
echo 133 2 | psxy -R -J -O -K -Sc1.4i -Gwhite >> GMT_App_K_3.ps  
psbasemap -R -J -O -K -Tm133/2/1i::+45/10/5 --HEADER_FONT_SIZE=12p --TICK_LENGTH=0.05i \  
        --ANNOT_FONT_SIZE_SECONDARY=8p >> GMT_App_K_3.ps  
getrect 100 | psxy -R -J -O -Wthicker -L -A >> GMT_App_K_3.ps
__________________________________________________________________________________________________

PIC

Figure K.3: Map using the intermediate resolution coastline data. We have added a compass rose just because we have the power to do so.

K.4.4 The high resolution (-Dh)

The relentless zooming continues! Now, the edges of the map are all 100 km true distance from the projection center. We step up to the high resolution data set as it is needed to accurately portray the detailed geographic features within the region. Because of the small scale we only ignore features less than 1 km2 in area. The high resolution database has undergone minor decimation and simplification by the DP-routine: The combined file size of the coastlines, rivers, and borders now swells to 12.3 Mbytes. The map and the final outline box are generated by these commands:

_________________________________________________________________________________

pscoast ‘getbox -JE130.35/-0.2/3.5i 100‘ -J -P -Dh -A1 \  
        -Glightgray -Wthinnest -N1/thinnest,- -B30mg10mWSne -K > GMT_App_K_4.ps  
getrect 20 | psxy -R -J -O -Wthicker -L -A >> GMT_App_K_4.ps
__________________________________________________________________________________________________

PIC

Figure K.4: Map using the high resolution coastline data.

K.4.5 The full resolution (-Df)

We now arrive at our final plot, which shows a detailed view of the western side of the small island of Waigeo. The map area is approximately 40 by 40 km. We call upon the full resolution data set to portray the richness of geographic detail within this region; no features are ignored. The full resolution has undergone no decimation and it shows: The combined file size of the coastlines, rivers, and borders totals a (once considered hefty) 55.9 Mbytes. Our final map is reproduced by the single command:

_________________________________________________________________________________

pscoast ‘getbox -JE130.35/-0.2/3.5i 20‘ -J -P -Df \  
        -Glightgray -Wthinnest -N1/thinnest,- -B10mg2mWSne > GMT_App_K_5.ps
__________________________________________________________________________________________________

PIC

Figure K.5: Map using the full resolution coastline data.

We hope you will study these examples to enable you to make efficient and wise use of this vast data set.

Appendix L
GMT on non-UNIX platforms

L.1 Introduction

While GMT can be ported to non-UNIX systems such as Windows, it is also true that one of the strengths of GMT lies its symbiotic relationship with UNIX. We therefore recommend that GMT be installed in a POSIX-compliant UNIX environment such as traditional UNIX-systems, Linux, or Mac OS X. If abandoning your non-UNIX operating system is not an option, consider one of these solutions:

WINDOWS:
Choose among these four possibilities:

  1. Install GMT under Cygwin (A GNU port to Windows).
  2. Install GMT under SFU (Windows Services for UNIX); a free download from Microsoft50.
  3. Install GMT under DJGPP (another GNU port to Windows/DOS).
  4. Install GMT in Windows using Microsoft C/C++ or other compilers. Unlike the first three, this option will not provide you with any UNIX tools so you will be limited to what you can do with DOS batch files.

L.2 Cygwin and GMT

Because GMT works best in conjugation with UNIX tools we suggest you install GMT using the Cygwin product from Cygnus (now assimilated by Redhat, Inc.). This free version works on any Windows version and it comes with both the Bourne Again shell bash and the tcsh. You also have access to most standard GNU development tools such as compilers and text processing tools (awk, grep, sed, etc.). Note that executables prepared for Windows will also run under Cygwin.

Follow the instructions on the Cygwin page51 on how to install the package; note you must explicitly add all the development tool packages (e.g., gcc etc) as the basic installation does not include them by default. Once you are up and running under Cygwin, you may install GMT the same way you do under any other UNIX platform by either running the automated install via install_gmt4.sh or manually running configure first, then type make all. If you install via the web form, make sure you save the parameter file without DOS CR/LF endings. Use dos2unix to get rid of those if need be.

Finally, from Cygwin’s User Guide: By default, no Cygwin program can allocate more than 384 MB of memory (program and data). You should not need to change this default in most circumstances. However, if you need to use more real or virtual memory in your machine you may add an entry in either the HKEY_LOCAL_MACHINE (to change the limit for all users) or HKEY_CURRENT_USER (for just the current user) section of the registry. Add the DWORD value heap_chunk_in_mb and set it to the desired memory limit in decimal Mb. It is preferred to do this in Cygwin using the regtool program included in the Cygwin package. (For more information about regtool or the other Cygwin utilities, see the Section called Cygwin Utilities in Chapter 3 of the Cygwin’s User Guide or use the help option of each utility.) You should always be careful when using regtool since damaging your system registry can result in an unusable system. This example sets the local machine memory limit to 1024 Mb:

regtool -i set /HKLM/Software/Cygnus\ Solutions/Cygwin/heap_chunk_in_mb 1024  
regtool -v list /HKLM/Software/Cygnus\ Solutions/Cygwin

For more installation details see the general README file.

L.3 SFU and GMT

SFU52 is also similar to Cygwin in that it provides precompiled UNIX tools for DOS/WIN32, including the sh and csh shells.

L.4 DJGPP and GMT

DJGPP53 is similar to Cygwin in that it provides precompiled UNIX tools for DOS/WIN32, including the bash shell. At the time of this writing we have not been successful in compiling netCDF in this environment. This is fully due to our limited understanding of the innards of the netCDF installation whose configure script did not work for us. As soon as this problem is overcome we expect a smooth install similar to that of Cygwin.

L.5 WIN32 and GMT

GMT will compile and install using the Microsoft Visual C/C++ compiler. We expect other WIN32 C compilers to give similar results. Since configure cannot be run you must manually rename gmt_notposix.h.in to gmt_notposix.h. The netCDF home page gives full information on how to compile and install netCDF; precompiled libraries are also available. At present we simply have a lame gmtinstall.bat file that compiles the entire GMT package, and gmtsuppl.bat which compiles most of the supplemental programs. If you just need to run GMT and do not want to mess with compilations, get the precompiled binaries from the GMT ftp sites.

L.6 OS/2 and GMT

GMT has been ported to OS/2 by Allen Cogbill, Los Alamos National Laboratory. One must have EMX installed in order to use the executables. All features that are present in the UNIX version of GMT are available in the OS/2 version. All executables may be obtained using links in the following document, which provides more detail on the port.

L.7 Mac OS and GMT

GMT will install directly under Mac OS X since it is fully Unix compliant.

Appendix M
Of colors and color legends

M.1 Built-in color palette tables

Figure M.1 shows each of the 22 built-in color palettes, stored in so-called cpt tables. The programs makecpt and grd2cpt are used to access these master cpt tables and translate/scale them to fit the user’s range of z-values. The top half of the color bars in the Figure shows the original color scale, which can be either discrete or continuous, though some (like globe) are a mix of the two. The bottom half the color bar are built by using makecpt -T-1/1/0.25, thus splitting the color scale into 8 discrete colors.


PIC

Figure M.1: The standard 22 cpt files supported by GMT.

M.2 Labeled and non-equidistant color legends

The use of color legends has already been introduced in Chapter 7 (examples 2, 16, and 17). Things become a bit more complicated when you want to label the legend with names for certain intervals (like geological time periods in the example below). To accomplish that, one should add a semi-colon and the label name at the end of a line in the cpt table and add the -L option to the psscale command that draws the color legend. This option also makes all intervals in the legend of equal length, even it the numerical values are not equally spaced.

Normally, the name labels are plotted at the lower end of the intervals. But by adding a gap amount (even when zero) to the -L option, they are centered. The example below also shows how to annotate ranges using -Li (in which case no name labels should appear in the cpt file), and how to switch the color bar around (by using a negative length).

_________________________________________________________________________________

#!/bin/bash  
#  
ps=GMT_App_M_2.ps  
 
gmtset ANNOT_FONT_SIZE 10p MEASURE_UNIT cm  
 
# Set up color palette with named annotations  
 
cat > ages.cpt <<END  
#COLOR_MODEL = RGB  
#  
0       197     0       255     23      197     0       255     ;Neogene  
23      81      0       255     66      81      0       255     ;Paleogene  
66      0       35      255     146     0       35      255     ;Cretaceous  
146     0       151     255     200     0       151     255     ;Jurassic  
200     0       255     244     251     0       255     244     ;Triassic  
251     0       255     127     299     0       255     127     ;Permian  
299     0       255     11      359     0       255     11      ;Carboniferous  
359     104     255     0       416     104     255     0       ;Devonian  
416     220     255     0       444     220     255     0       ;Silurian  
444     255     174     0       488     255     174     0       ;Ordovician  
488     255     58      0       542     255     58      0       ;Cambrian  
B       black  
F       white  
END  
 
# Top row, left to right. Using names.  
psscale -Ef -Cages.cpt  -D00/13/-8/0.5    -K         > $ps  
psscale -Ef -Cages.cpt  -D04/13/-8/0.5 -O -K -L     >> $ps  
psscale -Ef -Cages.cpt  -D08/13/-8/0.5 -O -K -L0.0  >> $ps  
psscale -Ef -Cages.cpt  -D12/13/-8/0.5 -O -K -L0.1  >> $ps  
psscale -Ef -Cages.cpt  -D16/13/+8/0.5 -O -K -L     >> $ps  
psscale -Ef -Cages.cpt  -D20/13/+8/0.5 -O -K -L0.1  >> $ps  
 
# Bottom row, left to right. Using numbers.  
sed ’s/;.*$//’ ages.cpt > years.cpt  
psscale -Ef -Cyears.cpt -D00/04/+8/0.5 -O -K        >> $ps  
psscale -Ef -Cyears.cpt -D04/04/-8/0.5 -O -K -L     >> $ps  
psscale -Ef -Cyears.cpt -D08/04/-8/0.5 -O -K -L0.0  >> $ps  
psscale -Ef -Cyears.cpt -D12/04/-8/0.5 -O -K -L0.1  >> $ps  
psscale -Ef -Cyears.cpt -D16/04/-8/0.5 -O -K -Li    >> $ps  
psscale -Ef -Cyears.cpt -D20/04/-8/0.5 -O    -Li0.1 >> $ps  
 
rm -f ages.cpt years.cpt
__________________________________________________________________________________________________

PIC

Figure M.2: The many forms of color legends created by psscale.

Appendix N
Custom Plot Symbols

GMT comes with several custom plot symbols ready to go. They are used in psxy and psxyz using the -Sk option. To make your own custom plot symbol, please follow the instructions given in the man pages of those two programs. The following is a plot of each symbol. Note that we only show the symbol outline and not any fill. Be aware that some symbols may have a hardwired fill or no-fill component. Also note that some symbols, in particular the geometric ones, duplicate what is already available as standard built-in symbols.


PIC

Figure N.1: Custom plot symbols supported by GMT.

Appendix O
Annotation of Contours and “Quoted Lines”

The GMT programs grdcontour (for data given as 2-dimensional grids) and pscontour (for x,y,z tables) allow for contouring of data sets, while psxy and psxyz can plot lines based on x,y- and x,y,z-tables, respectively. In both cases it may be necessary to attach labels to these lines. Clever or optimal placements of labels is a very difficult topic, and GMT provides several algorithms for this placement as well as complete freedom in specifying the attributes of the labels. Because of the richness of these choices we present this Appendix which summarizes the various options and gives several examples of their use.

O.1 Label Placement

While the previous GMT versions 1–3 allowed for a single algorithm that determined where labels would be placed, GMT 4 allows for five different algorithms. Furthermore, a new “symbol” option (-Sq for “quoted line”) has been added to psxy and psxyz and hence the new label placement mechanisms apply to those programs as well. The contouring programs expect the algorithm to be specified as arguments to -G while the line plotting programs expect the same arguments to follow -Sq. The information appended to these options is the same in both cases and is of the form [code]info. The five algorithms correspond to the five codes below (some codes will appear in both upper and lower case; they share the same algorithm but differ in some other ways). In what follows, the phrase “line segment” is taken to mean either a contour or a line to be labeled. The codes are:

d:
Full syntax is ddist[c|i|m|p][/frac]. Place labels according to the distance measured along the projected line on the map. Append the unit you want to measure distances in [Default is taken from MEASURE_UNIT]. Starting at the beginning of a line, place labels every dist increment of distance along the line. To ensure that closed lines whose total length is less than dist get annotated, we may append frac which will place the first label at the distance d = dist × frac from the start of a closed line (and every dist thereafter). If not given, frac defaults to 0.25.
D:
Full syntax is Ddist[d|e|k|m|n][/frac]. This option is similar to d except the original data must be referred to geographic coordinates (and a map projection must have been chosen) and actual Earth54 surface distances along the lines are considered. Append the unit you want to measure distances in; choose among degree, meter [Default], kilometer, statute miles, or nautical miles. Other aspects are similar to code d.
f:
Full syntax is ffix.d[/slop[c|i|m|p]]. Here, an ASCII file fix.d is given which must contain records whose first two columns hold the coordinates of points along the lines at which locations the labels should be placed. Labels will only be placed if the coordinates match the line coordinates to within a distance of slop (append unit or we use MEASURE_UNIT). The default slop is zero, meaning only exact coordinate matches will do.
l:
Full syntax is lline1[,line2[, ...]]. One or more straight line segments are specified separated by commas, and labels will be placed at the intersections between these lines and our line segments. Each line specification implies a start and stop point, each corresponding to a coordinate pair. These pairs can be regular coordinate pairs (i.e., longitude/latitude separated by a slash), or they can be two-character codes that refer to predetermined points relative to the map region. These codes are taken from the pstext justification keys [L|C|R][B|M|T] so that the first character determines the x-coordinate and the second determines the y-coordinate. In grdcontour, you can also use the two codes Z+ and Z- as shorthands for the location of the grid’s global maximum and minimum, respectively. For example, the line LT/RB is a diagonal from the upper left to the lower right map corner, while Z-/135W/15S is a line from the grid minimum to the point (135°W, 15°S).
L:
Same as l except we will treat the lines given as great circle start/stop coordinates and fill in the points between before looking for intersections.
n:
Full syntax is nnumber[/minlength[c|i|m|p]]. Place number of labels along each line regardless of total line length. The line is divided into number segments and the labels are placed at the centers of these segments. Optionally, you may give a minlength distance to ensure that no labels are placed closer than this distance to its neighbors.
N:
Full syntax is Nnumber[/minlength[c|i|m|p]]. Similar to code n but here labels are placed at the ends of each segment (for number 2). A special case arises for number = 1 when a single label will be placed according to the sign of number: -1 places one label justified at the start of the line, while +1 places one label justified at the end of the line.
x:
Full syntax is xcross.d. Here, an ASCII file cross.d is a multi-segment file whose lines will intersect our segment lines; labels will be placed at these intersections.
X:
Same as x except we treat the lines given as great circle start/stop coordinates and fill in the points between before looking for intersections.

Only one algorithm can be specified at any given time.

O.2 Label Attributes

Determining where to place labels is half the battle. The other half is to specify exactly what are the attributes of the labels. It turns out that there are quite a few possible attributes that we may want to control, hence understanding how to specify these attributes becomes important. In the contouring programs, one or more attributes may be appended to the -A option using the format +code[args] for each attribute, whereas for the line plotting programs these attributes are appended to the -Sq option following a colon (:) that separates the label codes from the placement algorithm. Several of the attributes do not apply to contours so we start off with listing those that apply universally. These codes are:

+a:
Controls the angle of the label relative to the angle of the line. Append n for normal to the line, give a fixed angle measured counter-clockwise relative to the horizontal. or append p for parallel to the line [Default]. If using grdcontour the latter option you may further append u or d to get annotations whose upper edge always face the next higher or lower contour line.
+c:
Surrounding each label is an imaginary label “textbox” which defines a region in which no segment lines should be visible. The initial box provides an exact fit to the enclosed text but clearance may be extended in both the horizontal and vertical directions (relative to the label baseline) by the given amounts. If these should be different amounts please separate them by a slash; otherwise the single value applies to both directions. Append the distance units of your choice (c|i|m|p), or give % to indicate that the clearance should be this fixed percentage of the label font size in use. The default is 15%.
+d:
Debug mode. This is useful when testing contour placement as it will draw the normally invisible helper lines and points in the label placement algorithms above.
+f:
Specifies the desired label font. See pstext for font names or numbers. The default font is given by ANNOT_FONT_PRIMARY.
+g:
Selects opaque rather than the default transparent textboxes. You may optionally append the color you want to fill the label boxes; the default is the same as PAGE_COLOR.
+j:
Selects the justification of the label relative to the placement points determined above. Normally this is center/mid justified (CM in pstext justification parlance) and this is indeed the default setting. Override by using this option and append another justification key code from [L|C|R][B|M|T]. Note for curved text (+v) only vertical justification will be affected.
+k:
Sets the color of the text labels, which otherwise defaults to that given by COLOR_BACKGROUND.
+o:
Request a rounded, rectangular label box shape; the default is rectangular. This is only manifested if the box is filled or outlined, neither of which is implied by this option alone (see +g and +p). As this option only applies to straight text, it is ignored if +v is given.
+p:
Selects the drawing of the label box outline; append your preferred pen unless you want the default GMT pen [0.25p,black].
+r:
Do not place labels at points along the line whose local radius of curvature falls below the given threshold value. Append the radius unit of your choice (c|i|m|p) [Default is 0].
+s:
Change the font size of the labels, which by default is 9 points.
+u:
Append the chosen unit to the label. Normally a space will separate the label and the unit. If you want to close this gap, append a unit that begins with a hyphen (–). If you are contouring with grdcontour and you specify this option without appending a unit, the unit will be taken from the z-unit attribute of the grid header.
+v:
Place curved labels that follow the wiggles of the line segments. This is especially useful if the labels are long relative to the length-scale of the wiggles. The default places labels on an invisible straight line at the angle determined.
+w:
The angle of the line at the point of straight label placement is calculated by a least-squares fit to the width closest points. If not specified, width defaults to 10.
+=:
Similar in most regards to +u but applies instead to a label prefix which you must append.

For contours, the label will be the value of the contour (possibly modified by +u or +=). However, for quoted lines other options apply:

+l:
Append a fixed label that will be placed at all label locations. If the label contains spaces you must place it inside matching quotes.
+L:
Append a code flag that will determine the label. Available codes are:
+Lh:
Take the label from the current multi-segment header (hence it is assumed that the input line segments are given in the multi-segment file format; if not we pick the single label from the file’s header record). We first scan the header for an embedded -Llabel option; if none is found we instead use the first word following the segment marker [ >].
+Ld:
Take the Cartesian plot distances along the line as the label; append c|i|m|p as the unit [Default is MEASURE_UNIT]. The label will be formatted according to the D_FORMAT string, unless label placement was determined from map distances along the segment lines, in which case we determine the appropriate format from the distance value itself.
+LD:
Calculate actual Earth surface distances and use the distance at the label placement point as the label; append d|e|k|m|n to specify the unit [If not given we default to degrees, unless label placement was determined from map distances along the segment lines, in which case we use the same unit specified for that algorithm]. Requires a map projection to be used.
+Lf:
Use all text after the 2nd column in the fixed label location file fix.d as labels. This choice obviously requires the fixed label location algorithm (code f) to be in effect.
+Ln:
Use the running number of the current multi-segment as label.
+LN:
Use a slash-separated combination of the current file number and the current multi-segment number as label.
+Lx:
As h but use the multi-segment headers in the cross.d file instead. This choice obviously requires the crossing segments location algorithm (code x|X) to be in effect.

O.3 Examples of Contour Label Placement

We will demonstrate the use of these options with a few simple examples. First, we will contour a subset of the global geoid data used in GMT Example 01; the region selected encompasses the world’s strongest “geoid dipole”: the Indian Low and the New Guinea High.

O.3.1 Equidistant labels

Our first example uses the default placement algorithm. Because of the size of the map we request contour labels every 1.5 inches along the lines:

_________________________________________________________________________________

pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_1.ps  
grdcontour geoid.nc -J -O -B20f10WSne -C10 -A20+s8 -Gd1.5i -S10 -T:LH >> GMT_App_O_1.ps
__________________________________________________________________________________________________

As seen in Figure O.1, the contours are placed rather arbitrary. The string of contours for -40 to 60 align well but that is a fortuitous consequence of reaching the 1.5 inch distance from the start at the bottom of the map.


PIC

Figure O.1: Equidistant contour label placement with -Gd, the only algorithm available in previous GMT versions.

O.3.2 Fixed number of labels

We now exercise the option for specifying exactly how many labels each contour line should have:

_________________________________________________________________________________

pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_2.ps  
grdcontour geoid.nc -J -O -B20f10WSne -C10 -A20+s8 -Gn1/1i -S10 -T:LH >> GMT_App_O_2.ps
__________________________________________________________________________________________________

By selecting only one label per contour and requiring that labels only be placed on contour lines whose length exceed 1 inch, we achieve the effect shown in Figure O.2.


PIC

Figure O.2: Placing one label per contour that exceed 1 inch in length, centered on the segment with -Gn.

O.3.3 Prescribed label placements

Here, we specify four points where we would like contour labels to be placed. Our points are not exactly on the contour lines so we give a nonzero “slop” to be used in the distance calculations: The point on the contour closest to our fixed points and within the given maximum distance will host the label.

_________________________________________________________________________________

cat << EOF > fix.d  
80      -8.5  
55      -7.5  
102     0  
130     10.5  
EOF  
pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_3.ps  
grdcontour geoid.nc -J -O -B20f10WSne -C10 -A20+d+s8 -Gffix.d/0.1i -S10 -T:LH >> GMT_App_O_3.ps
__________________________________________________________________________________________________

The angle of the label is evaluated from the contour line geometry, and the final result is shown in Figure O.3.


PIC

Figure O.3: Four labels are positioned on the points along the contours that are closest to the locations given in the file fix.d in the -Gf option.

To aid in understanding the algorithm we chose to specify “debug” mode (+d) which placed a small circle at each of the fixed points.

O.3.4 Label placement at simple line intersections

Often, it will suffice to place contours at the imaginary intersections between the contour lines and a well-placed straight line segment. The -Gl or -GL algorithms work well in those cases:

_________________________________________________________________________________

pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_4.ps  
grdcontour geoid.nc -J -O -B20f10WSne -C10 -A20+d+s8 -GLZ-/Z+ -S10 -T:LH >> GMT_App_O_4.ps
__________________________________________________________________________________________________

The obvious choice in this example is to specify a great circle between the high and the low, thus placing all labels between these extrema.


PIC

Figure O.4: Labels are placed at the intersections between contours and the great circle specified in the -GL option.

The thin debug line in Figure O.4 shows the great circle and the intersections where labels are plotted. Note that any number of such lines could be specified; here we are content with just one.

O.3.5 Label placement at general line intersections

If (1) the number of intersecting straight line segments needed to pick the desired label positions becomes too large to be given conveniently on the command line, or (2) we have another data set or lines whose intersections we wish to use, the general crossing algorithm makes more sense:

_________________________________________________________________________________

pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_5.ps  
grdcontour geoid.nc -J -O -B20f10WSne -C10 -A20+d+s8 -GXcross.d -S10 -T:LH >> GMT_App_O_5.ps
__________________________________________________________________________________________________

PIC

Figure O.5: Labels are placed at the intersections between contours and the multi-segment lines specified in the -GX option.

In this case, we have created three strands of lines whose intersections with the contours define the label placements, presented in Figure O.5.

O.4 Examples of Label Attributes

We will now demonstrate some of the ways to play with the label attributes. To do so we will use psxy on a great-circle line connecting the geoid extrema, along which we have sampled the ETOPO5 relief data set. The file transect.d thus contains lon, lat, dist, geoid, relief, with distances in km.

O.4.1 Label placement by along-track distances, 1

This example will change the orientation of labels from along-track to across-track, and surrounds the labels with an opaque, outlined textbox so that the label is more readable. We choose the place the labels every 1000 km along the line and use that distance as the label. The labels are placed normal to the line:

_________________________________________________________________________________

pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_6.ps  
grdcontour geoid.nc -J -O -K -B20f10WSne -C10 -A20+d+s8 -Gl50/10S/160/10S -S10 \  
        -T:’-+’ >> GMT_App_O_6.ps  
psxy -R -J -O -SqD1000k:+g+LD+an+p -Wthick transect.d >> GMT_App_O_6.ps
__________________________________________________________________________________________________

PIC

Figure O.6: Labels attributes are controlled with the arguments to the -Sq option.

The composite illustration in Figure O.6 shows the new effects. Note that the line connecting the extrema does not end exactly at the ‘-’ and ‘+’ symbols. This is because the placements of those symbols are based on the mean coordinates of the contour and not the locations of the (local or global) extrema.

O.4.2 Label placement by along-track distances, 2

A small variation on this theme is to place the labels parallel to the line, use spherical degrees for placement, append the degree symbol as a unit for the labels, choose a rounded rectangular textbox, and inverse-video the label:

_________________________________________________________________________________

pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_7.ps  
grdcontour geoid.nc -J -O -K -B20f10WSne -C10 -A20+d+um+s8 -Gl50/10S/160/10S -S10 \  
        -T:-+ >> GMT_App_O_7.ps  
psxy -R -J -O -SqD15d:+gblack+kwhite+Ld+o+u-\\260 -Wthick transect.d >> GMT_App_O_7.ps
__________________________________________________________________________________________________

The output is presented as Figure O.7.


PIC

Figure O.7: Another label attribute example.

O.4.3 Using a different data set for labels

In the next example we will use the bathymetry values along the transect as our label, with placement determined by the distance along track. We choose to place labels every 1500 km. To do this we need to pull out those records whose distances are multiples of 1500 km and create a “fixed points” file that can be used to place labels and specify the labels. This is done with awk.

_________________________________________________________________________________

awk ’{if (NR > 1 && ($3 % 1500) == 0) print $1, $2, int($5)}’ transect.d > fix2.d  
pscoast -R50/160/-15/15 -JM5.3i -Glightgray -A500 -K -P > GMT_App_O_8.ps  
grdcontour geoid.nc -J -O -K -B20f10WSne -C10 -A20+d+um+s8 -Gl50/10S/160/10S -S10 \  
        -T:-+ >> GMT_App_O_8.ps  
psxy -R -J -O -Sqffix2.d:+g+an+p+Lf+um+s8 -Wthick transect.d >> GMT_App_O_8.ps
__________________________________________________________________________________________________

The output is presented as Figure O.8.


PIC

Figure O.8: Labels based on another data set (here bathymetry) while the placement is based on distances.

O.5 Putting it all together

Finally, we will make a more complex composite illustration that uses several of the label placement and label attribute settings discussed in the previous sections. We make a map showing the tsunami travel times (in hours) from a hypothetical catastrophic landslide in the Canary Islands55. We lay down a color map based on the travel times and the shape of the seafloor, and travel time contours with curved labels as well as a few quoted lines. The final script is

_________________________________________________________________________________

R=-R-85/5/10/55  
grdgradient topo5.nc -Nt1 -A45 -Gtopo5_int.nc  
gmtset PLOT_DEGREE_FORMAT ddd:mm:ssF ANNOT_FONT_SIZE_PRIMARY +9p  
project -E74W/41N -C17W/28N -G10 -Q > great_NY_Canaries.d  
project -E74W/41N -C2.33/48.87N -G100 -Q > great_NY_Paris.d  
km=‘echo 17W 28N | mapproject -G74W/41N/k -fg --D_FORMAT=%.0f | cut -f3‘  
cat << EOF > ttt.cpt  
0       lightred        3       lightred  
3       lightyellow     6       lightyellow  
6       lightgreen      100     lightgreen  
EOF  
grdimage -Sc/1 ttt_atl.nc -Itopo5_int.nc -Cttt.cpt $R -JM5.3i -P -K > GMT_App_O_9.ps  
grdcontour ttt_atl.nc -R -J -O -K -C0.5 -A1+u~hour~+v+s8+f17 -GL80W/31N/17W/26N,17W/28N/17W/50N \  
        -S2 >> GMT_App_O_9.ps  
psxy -R -J -Wfatter,white great_NY_Canaries.d -O -K  >> GMT_App_O_9.ps  
pscoast -R -J -B20f5:.~Tsunami travel times from the Canaries~:WSne -N1/thick -O -K -Glightgray \  
        -Wfaint -A500 >> GMT_App_O_9.ps  
gmtconvert great_NY_*.d -E | psxy -R -J -O -K -Sa0.15i -Gred -Wthin >> GMT_App_O_9.ps  
psxy -R -J -Wthick great_NY_Canaries.d -O -K \  
        -Sqn1:+f6+s8+l~Distance Canaries to New York = $km km~+ap+v >> GMT_App_O_9.ps  
psxy -R -J great_NY_Paris.d -O -K -Sc0.08c -Gblack >> GMT_App_O_9.ps  
psxy -R -J -Wthinner great_NY_Paris.d -O -K -SqD1000k:+an+o+gblue+kwhite+LDk+s7+f1 >> GMT_App_O_9.ps  
cat << EOF | pstext -R -J -O -K -Wwhite,Othin -Dj0.1i/0.1i >> GMT_App_O_9.ps  
74W     41N     8       0       17      RT      New York  
2.33E   48.87N  8       0       17      CT      Paris  
17W     28N     8       0       17      CT      Canaries  
EOF  
psxy -R -J -O /dev/null >> GMT_App_O_9.ps
__________________________________________________________________________________________________

with the complete illustration presented as Figure O.9.


PIC

Figure O.9: Tsunami travel times from the Canary Islands to places in the Atlantic, in particular New York. Should a catastrophic landslide occur it is possible that New York will experience a large tsunami about 8 hours after the event.

Appendix P
Special Operations

P.1 Running GMT in isolation mode

In Chapter 4 it is described how GMT creates several (temporary) files to communicate between the different commands that make up the script that finally creates a plot. Among those files are:

.gmtdefaults4.
This file covers about 100 different settings that influence the layout of your plot, from font sizes to tick lengths and date formats (See Section 4.2). Those settings can be altered by editing the file, or by running the gmtset command. A problem may arise when those settings are changed half-way through the script: the next time you run the script it will start with the modified settings and hence might alter your scripts results. It is therefore often necessary to revert to the original .gmtdefaults4 file. Isolation mode avoids that issue.
.gmtcommands4.
This file is created to communicate the command line history from one command to the next (Section 4.5) so that shorthands like -R or -J can be used once it has been set in a previous GMT command. The existence of this file makes if impossible to run two GMT scripts simultaneously in the same directory, since those .gmtcommand4 files may clash (contain different histories) and adversely affect the results of both scripts.
.gmt_bb_info.
This file contains the information about the BoundingBox (Section C.1) of the PostScript output. This information too has to be transferred from one GMT command to the next in a script. Again, running two commands simultaneously in the same directory may have disastrous effects on that file.

A cure to all these woes is the isolation mode introduced in GMT version 4.2.2. This mode allows you to run a GMT script without leaving any traces other than the resulting PostScript or data files, and not altering the .gmtdefaults4 or .gmtcommands4 files. Those files will be placed in a temporary directory instead. And if properly set up, this temporary directory will only be used by a single script, even if another GMT script is running simultaneously. This also provides the opportunity to create any other temporary files that the script might create in the same directory.

The example below shows how isolation mode works.

_________________________________________________________________________________

#!/bin/bash  
#               GMT Appendix P, example 1  
#  
# Purpose:      Illustrates the use of isolation mode  
# GMT progs:    gmtset, grdimage, grdmath, makecpt, pscoast  
# Unix progs:   mktemp, rm  
#  
ps=GMT_App_P_1.ps  
 
# Create a temporary directory. $GMT_TMPDIR will be set to its pathname.  
# XXXXXX is replaced by a unique random combination of characters.  
export GMT_TMPDIR=‘mktemp -d /tmp/gmt.XXXXXX‘  
 
# These settings will be local to this script only since it writes to  
# $GMT_TMPDIR/.gmtdefaults4  
gmtset COLOR_MODEL rgb ANNOT_FONT_SIZE_PRIMARY 14p  
 
# Make grid file and color map in temporary directory  
grdmath -Rd -I1 Y = $GMT_TMPDIR/lat.nc  
makecpt -Crainbow -T-90/90/60 -Z > $GMT_TMPDIR/lat.cpt  
 
# The grdimage command creates the history file $GMT_TMPDIR/.gmtcommands4  
grdimage $GMT_TMPDIR/lat.nc -Sl -JK6.5i -C$GMT_TMPDIR/lat.cpt -P -K > $ps  
pscoast -R -J -O -Dc -A5000 -Gwhite -B60g30/30g30 >> $ps  
 
# Clean up all temporary files and the temporary directory  
rm -rf $GMT_TMPDIR
__________________________________________________________________________________________________

PIC

Figure P.1: Example created in isolation mode

The files .gmtdefaults4 and .gmtcommands4 are automatically created in the temporary directory $GMT_TMPDIR. The script is also adjusted such that the temporary grid file lat.nc and colormap lat.cpt are created in that directory as well. To make things even more easy, GMT now provides a set of handy shell functions in gmt_shell_functions.sh: simply include that file in the script and the creation and the removal of the temporary directory is reduced to a single command.

_________________________________________________________________________________

#!/bin/bash  
#               GMT Appendix P, example 2  
#  
# Purpose:      Illustrates the use of isolation mode  
# GMT progs:    gmtset, grdimage, grdmath, makecpt, pscoast  
# GMT funcs:    gmt_init_tmpdir, gmt_remove_tmpdir  
#  
ps=GMT_App_P_2.ps  
 
# Make GMT shell functions accessible the the script  
 
# Create a temporary directory. $GMT_TMPDIR will be set to its pathname.  
gmt_init_tmpdir  
 
# These settings will be local to this script only since it writes to  
# $GMT_TMPDIR/.gmtdefaults4  
gmtset ANNOT_FONT_SIZE_PRIMARY 14p  
 
# Make grid file and color map in temporary directory  
grdmath -Rd -I1 Y = $GMT_TMPDIR/lat.nc  
makecpt -Crainbow -T-90/90/60 -Z > $GMT_TMPDIR/lat.cpt  
 
# The grdimage command creates the history file $GMT_TMPDIR/.gmtcommands4  
grdimage $GMT_TMPDIR/lat.nc -Sl -JK6.5i -C$GMT_TMPDIR/lat.cpt -P -K > $ps  
pscoast -R -J -O -Dc -A5000 -Gwhite -B60g30/30g30 >> $ps  
 
# Clean up all temporary files and the temporary directory  
gmt_remove_tmpdir
__________________________________________________________________________________________________

P.2 Using both GMT 3 and 4

We encourage all GMT users to start using version 4 immediately; it has been tested extensively by the GMT team and has benefitted from bug reports for the 3.4.x versions. Users who still worry about the new version breaking things may install GMT 3.4.x versions and 4.x and use our utility gmtswitch to select their current version should the need to switch arises. You will find gmtswitch in the top-level GMT4.x directory; install as explained below.

Because GMT 4.x is backwards compatible with the 3.4.x series yet maintains its parameters and history in separate hidden files (e.g., .gmtdefaults4 versus .gmtdefaults) it is possible to install and use both versions on the same workstation. To simplify such setups we supply the utility gmtswitch which simplifies switching back and forth between any number of installed GMT 3-versions and GMT 4.x. Place the gmtswitch Bourne shell script in your general executable path (not in one of the GMT bin directories) and run it after you have finished installing all GMT versions of interest. The first time you run gmtswitch it will try to find all the available versions installed on your file system. The versions found will be listed in the file .gmtversions in your home directory; each line is the full path to a GMT root directory (e.g., /usr/local/GMT3.4.2). You may manually add or remove entries there at any time. You are then instructed to make two changes to your environment (the details are shell-dependent but explained by gmtswitch):

  1. gmtswitch creates and maintains a symbolic link this_gmt in your home directory that will point to a directory with one of the installed GMT versions.
  2. Make sure $HOME/this_gmt/bin is in your executable PATH.

Make those edits, logout, and log and back in again. The next time you run gmtswitch you will be able to switch between versions. Typing gmtswitch with no argument will list the available versions in a numerical menu and prompt you to choose one, whereas gmtswitch version will immediately switch to that version (version must be a piece of unique text making up the full path to a version, e.g., 3.4.2). If you use tcsh or csh you may have to type “rehash” to initiate the path changes.

1Version 1.0 was then informally released at the Lamont-Doherty Earth Observatory.

2Use standard UNIX tools such as awk or perl to reformat files should your date and clock components reside in separate columns.

3Used to color the background, foreground, and Not-a-Number areas.

4See GNU General Public License (www.gnu.org/copyleft/gpl.html) for terms on redistribution and modifications.

5The tools can also be installed on other platforms (see Appendix L).

6One public-domain RIP is ghostscript, available from www.gnu.org.

7Programs now also allow for fast, binary multicolumn file i/o.

8While the netCDF format is the default, other formats are also possible, including user-defined formats.

9PostScript definition. In the typesetting industry a slightly different definition of point (1/72.27 inch) is used.

10Choose between SI and US default units by modifying gmt_setup.conf in the GMT share directory.

11To remain backwards compatible with GMT 3.4.x we will also look for .gmtdefaults but only if .gmtdefaults4 cannot be found.

12The Gregorian Calendar is a revision of the Julian Calendar which was instituted in a papal bull by Pope Gregory XIII in 1582. The reason for the calendar change was to correct for drift in the dates of significant religious observations (primarily Easter) and to prevent further drift in the dates. The important effects of the change were (a) Drop 10 days from October 1582 to realign the Vernal Equinox with 21 March, (b) change leap year selection so that not all years ending in “00” are leap years, and (c) change the beginning of the year to 1 January from 25 March. Adoption of the new calendar was essentially immediate within Catholic countries. In the Protestant countries, where papal authority was neither recognized not appreciated, adoption came more slowly. England finally adopted the new calendar in 1752, with eleven days removed from September. The additional day came because the old and new calendars disagreed on whether 1700 was a leap year, so the Julian calendar had to be adjusted by one more day.

13While UTM coordinates clearly refer to points on the Earth, in this context they are considered “other”. Thus, when we refer to “geographical” coordinates herein we imply longitude, latitude.

14However, it is suppressed when a 3-D view is selected.

15Please consult the man page for printf or any book on C .

16For historical reasons, the GMT Default is Landscape, see gmtdefaults to change this.

17Ensures that boundary annotations do not fall off the page.

18For an overview of color systems such as HSV, see Appendix I.

19Convert other graphics formats to Sun ras format using ImageMagick’s convert program.

20For CMYK the format obviously involves two extra columns.

21Snyder, J. P., 1987, Map Projections A Working Manual, U.S. Geological Survey Prof. Paper 1395.

22This is, however, not the shortest distance. It is given by the great circle connecting the two points.

23Robinson provided a table of y-coordinates for latitudes every 5°. To project values for intermediate latitudes one must interpolate the table. Different interpolants may result in slightly different maps. GMT uses the interpolant selected by the parameter INTERPOLANT in the .gmtdefaults4 file.

24These data are available on CD-ROM from NGDC (www.ngdc.noaa.gov).

25These data are available on CD-ROM from NGDC (www.ngdc.noaa.gov).

26See http://topex.ucsd.edu/marine_grav/mar_grav.html.

27You can also use the utility curl

28Pedants who wish to argue about the “other” arc going the long way should consider using it.

29You could also use img2mercgrd directly – your only option under DOS

30While QuickTime is free you must upgrade to QuickTime Pro (USD 30) to use the authoring functions.

31QuickTime Pro can do this, as can most video-editing programs.

32Simon.Cox@csiro.au

33For data bases, see http://topex.ucsd.edu/marine_grav/mar_grav.html.

34Walter.HF.Smith@noaa.gov

35Kurt.Feigl@cnes.fr

36patau@ipgp.jussieu.fr

37The ASCII MGD77 data are available on CD-ROM from NGDC (www.ngdc.noaa.gov).

38These data are available on CD-ROM from NGDC (www.ngdc.noaa.gov).

39Timothy.J.Henstock@soc.soton.ac.uk

40lloyd@must-have-coffee.gen.nz

41In contrast, regular GMT PostScript files simply have a %%BoundingBox that equal the size of the chosen paper.

42http://www.bzip.org

43If you chose SI units during the installation then the default encoding is ISOLatin1+, otherwise it is Standard+.

44Note, however, that the -Q option in grdimage will exercise a PostScript Level 3 feature called colormasking.

45R. Bracewell, The Fourier Transform and its Applications, McGraw-Hill, London, 444p., 1965.

46www.ngdc.noaa.gov

47Douglas, D.H., and T. K. Peucker, 1973, Algorithms for the reduction of the number of points required to represent a digitized line or its caricature, Canadian Cartographer, 10, 112–122.

48The full and high resolution files are in separate archives because of their size. Not all users may need these files as the intermediate data set is better than the data provided with version 2.1.4.

49If you need complete polygons in a simpler format, see the article on GSHHS (Wessel, P., and W. H. F. Smith, 1996, A Global, self-consistent, hierarchical, high-resolution shoreline database, J. Geophys. Res. 101, 8741–8743).

50Microsoft Services for UNIX is formerly known as Interix, in the distant past known as OpenNT.

51cygwin.com

52See www.microsoft.com/technet/interopmigration/unix/sfu for details.

53See www.gnu.org for details.

54or whatever planet we are dealing with.

55Travel times were calculated using Geoware’s travel time calculator, ttt; see (http://www.geoware-online.com)